Performance stat suggestion - RWR

Hi guys,
Here is a proposal for a new performance statistic, based on win rate. Bear with me here, after a couple ...

Hi guys,

Here is a proposal for a new performance statistic, based on win rate. Bear with me here, after a couple of paragraphs of preamble, it will get to the point.

First of all, why have a new statistic? Currently there is win rate (WR) and other statistics (WN7, WN8, EFF). Each of these can be used per tank or per group of tanks (overall, by tank-type, etc). Consider unadjusted win rate – it is especially useful per tank, where it gives the overall expected player contribution to win probability, given the various advantages they chose to use (premium ammo, platoons, mods, etc.). If averaged over groups of tanks it is much less useful, as it depends on the specific tanks and their rate of play. The other statistics are intended as measures of inherent player skill, compensating for average performance of the various tanks so as to control for a player’s tank-mix and their relative rates of play. While these may offer good measures of skill in some ways, they are hampered by complicated formula, and can incentivize play to maximize their value, but not team wins.

In the abstract, to measure skill you might assign players the same tank, same crew, same map, or perhaps they would randomly be assigned a one of a number of pre-configured tanks, etc. In reality, in addition to choosing their tank, the player outfits the tank, chooses crew, and may arrive in a platoon of good (or bad) players. To measure skill, the goal is then to control for the effects of these player choices, after having identified likely influences. Unfortunately, WOT does not provide data for most of these (tank/crew configuration, platoons, etc), although there is some information available through the post-game files and VBADDICT. This can be used to improve on the basic win percentage (as per WN7, WN8, EFF) as a measure of skill.

The new statistic can be called a relative win rate (RWR, following relative risk in finance). For a specific tank it is the average win percentage by the player, divided by the expected win rate, as available through VBADDICT. If RWR>1 then it is better than average performance, RWR<1 is worse.

A difficulty with RWR (shared with the WR) is that for small samples it will be highly variable, as it takes many 0s and 1s to measure percentages with precision. And the situation is worse for the RWR, as it is about twice as variable as the percentage it is based on (since 100/typical expected ~2) There are a number of ways to deal with this, including the simplest approach (as with WR right now) of doing nothing. Preferable would be to report it as an interval estimate, so that the variability is clear (because it still lacks precision, even after 100 battles).

RWRs can be combined across different tanks by averaging. Like relative risk, you would use a geometric mean, so log transform, then average, then back-transform. The small sample size problem is an issue here, since 0% and 100% are found in small samples. Simple approaches would be to wait a certain number of battles (30?), or to wait for the first win and loss before including a tank in the average. Also simple (and preferable I think) would be to calculate the win percentage as (y+1)/(n+2)*100%, where y is the win count and N is the total number of games (this has justifications in statistics; it stabilizes the estimate and performs better than the direct proportion in small samples).

A weighted average can be used to adjust the averages for the different frequencies that tanks are played on an account. So, on the log scale, term i in the sum is multiplied by n_i/N where n_i is the number of games in the tank and N is the number of games in the group (groups such as all tanks; just spgs; etc.).

As an example calculation, suppose a player has two tanks, Rudy and KV-5. They win 60 out of 100 in the Rudy and 45 out of 100 in the KV-5. From VBADDICT, the expected win rate of Rudy is 56% and that of KV-5 is 52%. Their RWR in the Rudy is 1.07 and in the KV-5 is 0.87 (no small sample size adjustment here for clarity). Combined over both it is 0.96. If instead there were only 20 games in the Rudy with 12 wins, then the individual tank RWRs are the same, but the combined is now 0.90, down-weighting the fewer Rudy games. The expression for this is RWR=exp((N1/N *log(RWR1)+N2/N*log(RWR2))), where N=N1+N2.

I ran a little simulation to check its performance on a larger size sample. Suppose the player has 50 tanks, and plays 10, 20, …, 400 games in each (total = 8200 games) where each of these tanks has (from VBADDICT, say) a respective 40%, 41%, …,.60% win rate. The player is more skilled than average and has relative win rate of 1.04 on each tank. The simulation gave an estimated combined relative win rate of 1.035. An approximate 95% confidence interval for this is (1.014,1.055).

Here is a plot of the RWRs of the individual tanks in this simulation showing the decreased variability with tank number (red line is 1.04)

Running the simulation 100 times gave a distribution of the combined (over all 40 tanks) RWR

so the point estimates are pretty close, with means and medians where expected.

For reporting purposes, since the first few digits are often the same, it may be more interesting to report as a percentage. For example RWR 1.038 is 3.8% or 0.95 is -5%.

In conclusion, strengths are that it is simple to understand, is an improvement over WR for measuring skill (as opposed to predicting match outcome), and that it only goes up when the player wins games (as opposed to by damage, spotting, etc.), thus focusing directly on team play. Measures of uncertainty such as confidence intervals can also be calculated ether per tank or for group of tanks. Some weaknesses are that it cannot control for unmeasured effects such as platoons etc, and it does not control for skill differences among players in different tank types (although this can probably be done using existing data). That said, given what it does do, it can act as a complement to the various stats already available (with their own strengths and weaknesses).

The problem with measuring an individual according to win rate is that WoT is a team-based competitive game. You can perform extremely well, but still lose because your team didn't pull their weight. I've lost count the number of Ace Tanker medals I've won in losses. I have several of them posted to WoTReplays. While individual performance CAN affect the outcome of a battle, it just isn't enough most of the time to effect a win, because, as I stated, win rate is determined overall by team performance, not individual.

A better measure of individual performance would be a combination of damage ratio (either your damage compared to your hit points, or your damage compared to damage taken, or both), kill ratio (kills per battle and kills vs deaths), firing accuracy (hits/shots fired and penetrations/hits) and survival rate (battles survived/battles played). Including win rate in its algorithm, as WN8 does, is a mistake, in my opinion, and actually hurts the standard of measure for individual performance.

You win as a team, and you lose as a team, both because of and despite individual performance.

For example, look at a battle I had recently. I did over 3.3k damage, and killed 8 enemies. There were three of us left. So, the battle was decided by the remaining players. A T71, Type 58, and a Lowe (who was capping) for the enemies, and an ELC, a Cromwell, a T71 and myself in a Hellcat for our team. I was capping, while the Cromwell defended my capping from the Type 58. The enemy Lowe had 20 seconds or so lead in capping, so I wasn't going to out-cap him in time. Rather than help me cap, the Cromwell decided to run off and try to reset the Lowe, after he and our T71 took care of the Type 58. He was killed by the enemy T71 on his way there.

The ELC was totally useless, both as a scout, and in the final moments of the battle. He was damaged already, and died immediately while trying to rush the Lowe for a reset.

Our T71 rushed off too, rather than help me cap. He managed to kill their T71, who was hanging around our base to defend the capping Lowe. Then, he went after the Lowe to reset so he could have time to restrategize, but his one shot at the Lowe bounced, and he was killed.

We lost because THEY failed to help me cap, despite my exemplary performance in that battle, and my suggestions to help me cap for the win. We lost because my remaining teammates wanted more damage and kills.

My performance rating should not be affected by the failures of others.

Another example is when your team is doing well as a whole, but you haven't done much personally. Then, while some of you are hunting down the rest of the enemy, some teammates decide to sit on enemy cap. You win, but you've got only a few hundred damage and no kills.

To conclude, because you can perform extremely well in a loss, and perform extremely poorly in a win, win rate is an untenable standard of indiviual performance in a team-based environment.

Fireholder1,

I read some mixed messages here - you say "We lost because my remaining teammates wanted more damage and kills" but then also say "A better measure of individual performance would be a combination of damage ratio (either your damage compared to your hit points, or your damage compared to damage taken, or both), kill ratio (kills per battle and kills vs deaths)"

In any case, I agree that what matters is contribution to team win, and that is what RWR measures, as it is only a function of the win percentage. Situations where you outperform (as measured by other stats) in losses or underperform in wins will always happen, but over many battles contributions to wins and team play are what is rewarded by RWR - it penalizes maximizing other stats if this is done at the expense of overall team performance.

It wasn't a 'mixed message'. Rather it was explicit. They sacrificed a win to try to get damage and/or kills, or the Defender medal, or they thought they could reset to guarantee a win. Whichever the case, they directly contributed to a loss, outside my performance.

There's a problem when it comes to this game regarding goals in a particular battle. Some people play 'for fun'. Some people play to win. Some people play to pad their stats. And some people play for a combination of all of those. However, even fewer will sacrifice their own stats to ensure a win for the team. There is so much that this game cannot measure, such as driving in front of a tracked teammate to cover him from enemy fire while he repairs. Or intentionally drawing enemy fire so a weaker teammate can flank.

A simple algorithm predicated on rate of victories is horribly inadequate to measure someone's performance, which is precisely why the WN system was created.

Interesting topic, thanx to ekia2008 for the idea.

From a programmatically view this should be not to much work to implement into Wot Numbers. Wot Numbers already have exp win rate per tank from the WN8 rating, and of course it has win rate for the player for all tanks per battle mode. The formula is not very complex, and calculations like this is easy to add into our framework.

The suggested RWR rating could be an interesting way of measuring / rating a players win performance. We all play to win, don't we? A good player will over time have higher win rate than a bad player. I am quite an average player myself, but I have improved my skills gradually and my win rate has also improved. It less fun playing a poor tank with limited possibility to have an impact on the result compared to driving a good / OP tank. I guess this formula would compensate for that using the expected win rate to normalize the result (if I understood it correctly).

The issues with using win rate at all as a parameter for rating player performance, as Fireholder1 has commented above, is a never ending discussion probably Here is some of my thoughts:

WN8 can be used to measure a player performance in a single battle (we use the WN8wrx formula for this where win rate is then set equal to exp win rate). I think WN8 does an all right job for this, would love to see assisted damage be included though. The suggested RWR rating cannot be used to measure performance for a single battle, as win rate has no meaning for a single battle. I think it is a bit off topic to discuss if the suggested RWR rating is better or could replace WN8, it is to different things.

A more interesting discussion is: would it be a better rating than the plain win rate we use today? Win rate today amongst players are normally 40-60, I guess most around 48-52. Will the RWR formula actually rate players differently, and will players understand it if the scale is -5% to +5%?

However, WN8 rating and the plain win rate seems to be the rating most players care about today. These ratings has been there for a while, and it will probably not be easy to add a new rating and get attention for it unless several WoT services starts using it. I would guess any rating not supported by XVM mod and the biggest WoT stat sites (like vBAddict and Wot Labs) will struggle achieving popularity. Wot Numbers is a rather small 'player' amongst all WoT statistical 'vendors' and we have limited influence.

Cannot draw any conclusion yet, it looks interesting though.

A last final tips: we tend to be quite kind regarding request from donators and patrons

I think it can work, as ekia stated, as a complementary portion of WoT Numbers, whereby individuals can use it for their own purposes. As a publicly available statistic, though, it's relatively useless as a measure of individual performance. I always shake my head whenever someone mentions a player's win rate, because it's so easily manipulated, and it's not a measure of that individual's performance, but a measure of the performance of every TEAM that player has been on.

fireholder1,- your objection seems to be that there is more and better data to measure performance in a match. Certainly there is more. But, if the point of the game is to win matches then this data offers only an indirect approach to measuring contribution to that. In contrast, the percentage wins is the obvious place to start as it measures what you are interested in. But of course it suffers from being just a 0 or a 1 at the end of the game. This is what you might call a bias-variance tradeoff. With the huge amounts of in game match data you can create good predictors of performance that are biased for measuring contribution to win, andas you increase the size of your sample they will become more precisely biased. In contrast, if you work with the win rate you are immediately estimating the thing you want, however, have to wait until moderate to large samples to do so precisely.

There is also one more thing - if you use in game data you can develop a statistic where the tail wags the dog - knowing the formula for the statistic players will try to change their playstyle to maximize the statistic at the expense of team performance. This cannot happen for the win rate and is another advantage of the approach.

The point of the game is not simply to win matches, however.

The point of the game has several parameters, such as damage, kills, assistance (spotting/tracking), capping, defending. Your individual performance in a battle is irrelevant to a win or a loss, which is why even WG rewards players who earn Battle Hero medals in a loss (Confederate, Defender, High Caliber, Patrol Duty, Scout, Steel Wall, Tank Sniper, Top Gun) the same XP and credits they'd have earned in a win.

A major problem in public team play are people who play with the mentality of 'win at all costs' as if it were a good thing. No. Public battles are designed for earning credits and XP so players can advance their tanks and crews. It's not just about winning. The only time 'win at all costs' works is in tournaments, Strongholds, Clan Wars, and eSports, where earning credits and XP is not the primary goal.

Win rate can most certainly be manipulated, and often is via platooning (including and especially 'platoon drops', where several platoons enter battle queue simultaneously, hoping and often succeeding in landing in the same battle and/or on the same team) and tank companies.

The example I gave earlier, of players sitting on cap to get a quick win, is not 'team play'. They're in it just to win and get the bonuses for the cap points. They'd have served their team better by helping the rest of their team survive, but instead, they sit on cap while some of their teammates get stomped because they had no support.

Selfishness is a huge problem in this game, and measuring performance based on win rate won't change that a bit.

FYI, the game already records your win rate per tank, and win rate over all your battles.

And once again, win rate is not a measure of YOUR performance, but that of the TEAMS you have been on.

badbutton, - Thank you for your comments. Yes the formula would compensate for the quality of the tank as you say. I also agree that RWR and WR are different things from EFF, WN7, and WN8, for some of the reasons mentioned. I brought the topic up because personally I like the WR and was thinking about how it could be adjusted to measure player skill (controlling for tank type and frequncy of play). Also, RWR supplements the plain WR in this way and both (to me) are interesting, as they do different things.

You mention how to report it, and will players understand it if the scale is -5% to +5%. This is a good question, I suppose one way to look at it is that most players may not understand WN7, WN8 either. The obvious alternative is the ratio, but it seem to have few too many digits.

fireholder,

For most people I think it is fair to say that the main point of WOT is winning the match, they may have other goals , but that is numero uno.

Also, over many matches, WR *is* a measure of your performance. Definitely not in 1 or 2 mathces though!

Finally (last thing I see in your emails), yes WR and WRR are affected by platoons, premium ammo, etc, as are the other stats.

No, win rate is not a measure of YOUR performance. It's a measure of the performance of the teams you have been on. NOBODY wins an entire battle alone. Nobody. Pretending otherwise is self-delusory.

You are having an argument with basic probability theory. Take a slightly unfair coin biased to heads and flip it once - will it come up heads on any given toss? No. Toss it many times and you are guaranteed that the percentage of heads will come arbitrarily close to the true probability.

Take two teams stocked with equally average players except for a pro player on one. Which team would you bet on? Would the team with the pro-player have a 50% chance to win over the long run?

False dichotomy; there's a third option, and it's called 'draw'.

I think you are about run out of good arguments in this discussion Fireholder1

It's allowed to have different opinions about these matters, but perhaps we should focus the discussion around the suggested RWR formula and how it could work out. I think we have come to an end in the discussion if win rate at all is usable or not as a statistical measure. We can probably agree that we disagree.

For those who don't like win rate at all as a parameter to measure progress and performance don't really need to be concerned about this. I am tempted to test it out, but as any other value in Wot Numbers - you can decide yourself if you want to show it or not

I never said he couldn't have an opinion, nor did I try to dissuade him from voicing it. I'm simply disagreeing with him, using the fact that win rate is a measure of team performance, not individual. His poor analogy of a coin toss doesn't fit, because as I stated, the game has three options, not two.

If he wants to kee insisting something is true, when it isn't, that's his prerogative. Are you suggesting I refrain from disagreeing with people?

I already suggested that it wouldn't be a bad addition to WoT Numbers. I've only disagreed with his claim that it's a measure of individual performance, that is all.

Your contribution to discussions is always welcome Fireholder1, it's perfectly OK to bring good arguments to the table and see things from different perspectives.

It might be that you misunderstood the example above with the coin flipping. Of course, flipping a coin has only two possible outcomes: head or tail, while a battle has three outcomes: victory, draw or defeat. That is just the number of possible results, like rolling a dice has six possible outcomes. In statistics this makes no difference in regards to calculate probabilty. I would absolutely say that the analogy with the coin flipping is quite OK, you just have to put it into a statistical context to get it.

As with probabilities, it's also true that no matter what you calculate mathematically, real-world testing doesn't match the probability. It may come close, but it will never be spot on, because of chance. So, a coin has a 1 in 2 probability of landing on either side, but that does not mean it will land on either side evenly, regardless of how many tosses you test it with.

Ekia has clearly put some good thinking into his RWR algorithm, and I do think it's worth testing out for WoT Numbers, and for individuals to keep track for themselves. I would suggest, however, that it be altered to account for draws, rather than just wins and losses. This will make it more accurate. Not sure exactly how you'd wish to treat a draw within the algorithm, though.

Here is my first suggestion, calculatd for the tank view. Can you see if my calculations are correct according to your idea ekia2008?

The value is reported according to this suggestion from the inital idea description:

I added some coloring as well, using the standard colors like this:

<= 0 : Red / Orange / Yellow

> 0 : Green / Blue / Purple

About draws, in WOT a draw is a loss so there is no problem with this.

Badbutton - nice stats! Wonder what you get when you group those. The colors are a nice touch.

It is of course variable in small samples so it will be interesting to see how that plays out. This is true of the win rate too, of course, and to a lesser extent for the others (which even in large samples seem to show alot of variability across my tanks - perhaps it is just an old account). For the sample sizes you have the values should at least trend in some direction, assuming there is a trend.

When I programmed it I estimated the win rate as WR = (y+1)/(n+2)*100% to avoid any issues with zeroes.

Thanks for giving this a try. I will pitch something in to your fund in the next few days (although your paypal page is in Danish?)

Thanx for feedback. I did not use +1/+2 just the pure formula. I have a check for tanks with 0 battles, returning NULL = no rating.

I'll continue working on it for the home view, adding total calculations.

Sorry about the PayPal donate page, will change it to English language.

I have one question regarding your formula, Ekia.

Why are you adding 2 to the total battle count, instead of 1?

I guess that is just a simple way to avoid division by zero, he is adding 2 non existing battles with 50% win rate. It avoids division by zero and normalize the result when few battles are performed. The downside is that it makes the calculation not 100% mathematically correct.

I have selected to not use this, and just ignores RWR calculations for tanks without battles. This division by zero problem will anyway only happen after getting a new tank in the garage and before any battles is played with it.

Sounds good.

Here is the formula I have landed on:

WinCount and BattleCount is number of wins and total battles for a tank, or for a set of battles for a tank. WN8ExpWinRate is a 0-100 value presenting the expected win rate for a tank according to WN8 expected value, example: 52.5

When calculation is done for several tanks, or for a set of battles it will be run in a loop like this (pseudo code):

Here is the method: GetTankBattleInfo(<parameters>) explained:

The method will return a set of battle results per tank. It should return wins, tot battles and WN8 exp winrate for each tank separately in rows.

Parameters should specify battle mode, the WN8 expected values are calculated for random battles, so for other battle modes the expected values does not actually fit.

Parameters could spesify a spesific tank or a set of tanks like: tank type, tier or nation. It could also specify a set of battles like: battles today, latest week or lastest month.

A short news bulletin:

I have been in contact with Phalynx running vBAddict regarding support for RWR. He has looked into this thread and I have sent him som details about the calculations.

He is positive to it, and would like to support it. When and how this will be implemented in vBAddict is yet to be decided

That's awesome.

Why did you omit OP's logarithmic functions?

Oh, and a friendly FYI, your 'pseudocode' broke the formatting of your email notification, which resulted in all your HTML code showing up (the go to forum link, and the unsubscribe link).

Ahh, I guess my self-made notification solution did not manage to display all the strange formatting in the pseudo code comment. Sorry for that

I have to loop through the database to get tank battle info and tank exp winrate anyway, so just adding values up along the way and then use them in the generic formula was the most conveniant way for me. I guess my math knowlede comes to a limitation using logarithmic functions - so I had to simplify that part, I guess the end result will be the same (hopefully?)

You know coding, Ekia knows math; put your heads together and you can make it work.

A logarithm is simply the reverse of exponentiating a number. Basically, it's the number that is the exponent of one number (the base) in order to arrive at another number.

For example, 100 has a base factor of 10, so the logarithm (log for short) of base 10 for 100 = 2. 10 (base) raised to the exponent 2 (log) results in 100.

Or, the log of 32 with a base of 2 = 5. 2 (base) raised to the exponent 5 (log) = 32.

An extremely simple introduction to logarithms, but it covers the main idea. Base 2 (binary) logarithms are used in computer sciences.

I think what Ekia is attempting is something similar to an Exponential Moving Average, as it's used in finance. But I could be mistaken.

Thanks' for the math lesson, it was actually very good - even I understood it

The formula suggested by Ekia can be used to get a "total" weighted average out of a series of averages. For Wot Numbers this could work for tank statistics, as I already have total RWR per tank. However, it would not work for all scenarios. Some examples: RWR latest month or any other subset of recorded battles, or calculate RWR back in time by reversing stats according to players final stats subtracted recorded battles for a selected period.

The way I have built calculations for all ratings is by making only one final method for each rating calculation, and some more methods for gathering parameters. This makes it possible to reuse methods, and splits up the process for more manageble code. It also makes it possible for me to implemet quite complex calculations without using complex matematical funtions

About the formula, it is as BudButton suggests. It stabilizes the estimate and gets very close the the usual y/n as sample increases. It is also mathematically correct in a statistical way (posterior mean binomial random variable w/uniform prior), but it also can lead to more questions than necessary. Dropping these tanks should work too. The standard estimator (y/n) can also be problematic in the geometric mean if the player only has losses for some tanks (log 0 which in my software is -Inf and hence the group RWR is zero). To simplify my coding I used (y+1)/(n+2) in the geometric mean for that reason.

BadButton - thanks for sending it along to Phalynx. It will be interesting to see what he can do with it. Also, will take a look at your more recent posts later.

Thanks' for feedback ekia.

I have now implemented RWR for the home view as well, the rating is available in the total stats gadget and as separate gauge.

Will release this as a pilot version soon, and I hope it is OK if I add you as a test pilot ekia - so you can check if it all works out.

Here a screenshot of my stats. It will reveal me as a mediocre player (perhaps I should spend more time playing instead of programming). Added a gauge for my skirmishes battle performance as well, it helps when my clan team mates carry me

Just an idea regarding the usefulnes of WR or RWR as a measure of personal performance: Wot Numbers is actually tracking battle achievements, although currently not showing them in the battle results. That means it's possible to check if the player was a 'battle hero' on a battle that was lost, and in turn this can be used for weighting the RWR (or WR for that matter). It could for instance count as a netrual result instead of a loss. The point being that it would factor in the player's performance in lost battles where the he performed well :)

New pilot version 0.10.25 is released for test pilots with the new RWR rating implemented.

Ekia2008: to be able to download the pilot version you have to make sure that your forum user alias name is the same as your WoT ingame name.

I have created a Home View Setup with RWR rating gadgets included.

You have to use Wot Numbers 0.10.25 or later for this to work.

Nice dials for the stats. One idea in looking at it...

The idea of RWR (and also WN8,...) is to measure player skill, and current statistics (AFAIK) assume this fixed. But it goes upover time... The effects on performance stats are especially problematic if the player keeps their initial account and then battles from the learning period weighted equally as the later battles. A simple solution would be to drop the first X, where this is some number where most people are through the steepest part of the learning curve. Or, better, drop the first Y % of all games, to allow the "less skilled" games at the start to drop off. Not sure what a good value for Y would be (but pick a value (10%, 20%?) to start.

This would only be for the grouped stats. The individual tank calculations could include all games.

Will look at more emails later.

This logic is implemented for WN8 ratings at some sites, like noobmeter where you can see WN8 for latest battles (not sure what time range is, but probably lates days), last week, last 1000 battles and some more.

We have no such options within Wot Numbers (so far), but we have added charts for some values that illutrates the progress over time. It does of course not affect the end ratings value, but shows the player how the progress has been over time.

I gave your latest suggesting another thought ekia, it is actually not possible to implement this logic. WoT only reports players total stats per tank per battle mode. It is not possible to "remove" the first Y or Y% battle results.

When using a tool like Wot Numbers or vBAddict we fetch each battle results and stores this in our systems database. This makes it possible to track progress, and reverse stats by using players total stats and subtract single battle result stats. If you start using Wot Numbers after played X numbers of battles, stats for these X battles will not be available from any source.

That's too bad - it certainly limits these various stats. Is there other information to eliminate earlier battles? For example I look at my tank list and there are many that predate when I used your software and these can be identified by last battle stamp. If you progressively eliminate tanks by last played date to get your sum to a threshold that might do it. And any effect due to some tanks held for along time (or in play at the time of the last date) would probably be negligible.

Bad button - about the formula:

it has an extra "-1" in there, I think.

Also, for the pilot version - where is the download link? Thanks.

hello cmdrtrinity - interesting idea. It might work into future modifications or other stats. That said, for RWR I tried to avoid the within-battle information to keep things focused just on the win, as far any given match goes (plusthe within-game data requires much more effort to develop I think ).

I guess the formula is a bit misleading because in my method the WN8ExpWinRate is multiplied with battleCount when it reach the calculation, so it will be more correct to show the formula like this:

This could even be simplified like this

Example: you play 1000 battles, and win 540. Your Win rate is 54%. WN8 epected win rate is 52%, so you are a bit better than the average player according to WN8.

Insert into the formulas above:

This result says that you are 3.85% better than an average player according to WN8 exp values for win rate. To control this we migth calculate the players win rate that is the exp win rate + 3.85% better:

In other words, the "-1" has to be there to make this formula work.

Ah - I got confused. I think of RWR as the ratio. You are converting to the +/- percentage which is what is presented in the software as RWR.

About the WN8 expected - that is the expeced win rate with the tank, which is an input to the WN8 formula, right?

ALso, one other thing is that ideally the RWR (and all other player performance stats) would be calculated using just Random Battle data. Is it possible to subset to that?

Sorry for late reply, I missed this one

Now released in public version 0.11

Hi,

I finally got the pilot working (duh!) and had a look - it looks great (although my own overall RWR is -ve )

Does the "Tanks" tab allow display of the indiviual tank RWRs? In the setup I found this:

But not an RWR on the right panel to select. With the individual tanks RWR (and the inputs) I could reproduce your calculations for basic groupings, as a check. May take a week or two (extremely busy right now).

One other thing is that although overall I am negative, in the last 3 months I have been doing quite well (+476%) apparently. The calulations for "last three months" use the WOT Numbers database I bet. That would be worth a separate check (perhaps the individual tank last 3 month RWRs could be obtained in some way to see which ones are big or small).

To add columns you find it in the left side "Available Columns", and move it to the right: "Selected Columns".

To calculate RWR latest 3 months, month or week it uses the recorded battles from the battle view. If you only have recorded battles the last week - these three options will show the same value.

There is no option (at least yet) to show RWR per tank for a special period.