Source: http://forum.worldoftanks.eu/index.php?/topic/468409-a-1000-battle-study-on-xvm-win-chance-accuracy/
A 1000 Battle Study on XVM Win Chance Accuracy.
Author: Spitfeuer117
Published with author’s permission.
If you have XVM with statistics enabled in it, it will give a ”Chance to win” percentage for the battle. I went on a quest to find out whether you would actually win that battle that many times out of a hundred.
The only way to be sure would be to play the exact same battle thousands of times under the same conditions. Instead, I started to evaluate my win percentages under the battles with the same XVM predictions.
Out of 1000 battles, there were 598 wins, 390 losses and 12 draws. The average XVM prediction was 55%, which seems low compared to the 60% win rate.
By sorting the number of wins, losses and draws into the XVM prediction categories, the following histogram was produced:
By plotting the win rate in each category as function of the XVM prediction, we get the following figure. Only the categories with more than 20 battles (44%-66%) were taken into account. Also, draws were ignored as XVM does not take them into account in win chance calculation, i.e. two equal teams have both 50% predictions.
From the graph we can see that at 50% XVM gives a correct result, indicating that it calculated my skill (or my impact in the games) well. However, the higher win chance predictions are increasingly too low and the lower are increasingly too high.
XVM calculates the win chance by first calculating a skill value for each player by using a large number of statistical variables. It then calculates the skill of each team, Ka and Ke, recorded in xvm.log, by adding the player values together. The win chance is then calculated using the following formula:
P(win) = ((Ka / (Ka + Ke) – 0.5) * 1.5 + 0.5) * 100%
If P(win) > 95%, P(win) = 95%. [1]
Basically it scales the skill percentage with an arbitrary factor of 1.5. According to the figure above, the scaling factor should in fact be 2.3 times higher, 3.5, to produce more accurate results. Win chance as function of the skill percentage with the different scaling factors would look like this:
The correct way of determining the probability of a victory is very different however. If we assume that player’s effectiveness in a battle is the sum of infinitely many variables, it and therefore also the team’s effectiveness is normally distributed with an average of the skill value calculated by XVM for example. If we then assume that the winning team is the one with a higher effectiveness in the battle, the win probability is
P(win) = Φ((Ka – Ke) / s) * 100%
where Φ is the cumulative distribution function of the standard normal distribution and s the standard deviation in the teams’ skill difference. The standard deviation can be assumed constant and determined experimentally. A value of s = 30 seems to provide correct results. Figure below shows the shape of the function.
Note: Different horizontal axis. The use of skill difference instead of ratio brings low tier low skill battles closer to 50%, which makes sense. For example the figure below shows win chance with the normal distribution formula and XVM formulas with a scaling of 1.5 and 3.5 as function of enemy team skill, if allied team skill is 1.5 times higher.
I tested the normal distribution formula with the same method with 324 battles, and the following results were produced:
Although the number of battles was small, the normal distribution method provided roughly correct results. My win rate during the 324 battles was 61% instead of 60%, which is probably why the results are slightly higher than expected.
Conclusion
XVM is inaccurate. You win for example 27% of 40%ers, 50% of 50%ers and 73% of 60%ers.
[1] http://www.modxvm.com/en/faq/how-is-the-teams-chance-to-win-calculated/
Sooo, as I said and I will always say.. don’t look at the god damn win chances, they mean nothing.
No, use more accurate formulas.
They still means nothing on a single event. You cannot predict one even based on the statistics of a large collection.
All these formulas are not proper models, they lack variance distribution analysis for example. And you cannot do that on this type of scenario. All type of events based on the interaction of human beings need to be predicted with way different models that are not simple statistical inference. Its the same as with economy.. so many statistical models.. and they are all horrible and have bad results.
That is why when you use a pre election information on the vote intentions you cannot infer the chance of a single person that you pick on street to vote at a specific candidate. The amount of people that THINK that statistics work that way is astonishing. A population is not the same as a set of roll dices where the events are basically the same every time.
These data can only be used to predict the results of a series of matches not a single event. Single event predictions are only really doable in much simpler systems where there are not HUGE number of variables, some which are not even possible to evaluate before the match starts.
They don’t mean anything unless you get into the extremes.
When I see XVM say 5% win chance it;s usually pretty damn accurate.
When XVM says we have 95% chance to win, thats pretty accurate as well.
But when XVM is between 40 and 60%, anything can happen.
Dunno, having tracked my extremes, I tend to have found that 5% and 95% teams resulted in my winning 67% of them, both 5% and 95% having roughly the same margin of victories.
Once I had collected that data (lower sample size ~600 or so games), I had opted to never turn on win chance again, as winning more of my predicted terrible games, meant that it wasn’t correct. I didn’t even have to carry many of them.
I’ve seen quite a few kamikaze tankers trying to boost the accuracy of XVM on some lower percent matches.
they’re even so generous enough to chat the win-chance so non-XVMers can share the wisdom.
I have always viewed the XVM win percentages in reverse: XVM merely tells people what XVM thinks of our team’s players and the enemy team’s players. I can do the same thing just by looking at the list, counting greens, blues, and purples, and what tank they’re driving at the time.
Ultimately, I’m driving an SPG, so I just roll with it, do my part, and pray. :P
I look o n win rate to compare skill of my team to enemy team.
The article says the exact opposite to me: when XVM indicates a low chance to win, the real chance to win is even lower than the indication. When XVM indicates a high chance to win, the real chance to win is really even higher than the indication. This is actually an argument in favor of letting XVM win% dictate your behaviour in battle: for low winchances, you must play conservative to conserve HP so you can carry, while for high winchances you must #yolo to try and farm as much damage as possible.
To me this article demonstrates the need to shut down XVM by allowing players to set their stats private, or by disabling XVM’s ability to fetch player’s stats.
The article says that you should frantically spam/ping a lot more on the map in a low win-chance game, and less with a high win-chance. ;)
XVM can predict the TREND.. But it cannot give an accurate reading, becuse that is impossible on a system as complex as the interaction of 30 people within a 15 minute frame with possible combinations of tanks compositions , map placement, matching or not player style of people within a group etc that surpasses trillions of possible combinations.
So XVM can tell you if things LOOK bad or LOOK good. But is only a gross estimation.
> with possible combinations of tanks compositions , map placement, matching or not player style of people within a group etc that surpasses trillions of possible combinations.
Or if some players suddenly get a call during the match, their mom/wife drags them away from the computer, they just broke up with their girlfriend and their last 250 games (out of 10k) were absolutely miserable… the list goes on and on and on.
The statistics are complete bullshit as there are too many factors that they (have to) ignore.
In principle, that’s not at all impossible. Speaking as a professional data scientist, I can tell you that people model MUCH more complex situations, with useful results. Therefore, the complexity isn’t really the problem–in fact, this is, by the standards of statistics, a pretty well-defined system, with a relatively small number of variables (granted, there’s a lot of random noise–the “trillions of combinations”–but that doesn’t mean there are a lot of core variables).
The problem, instead, is lack of access to all of the data. We know each player’s WR in each tank, but what we’d like, ideally, is to know all the players involved in each battle (this would also get around the modeling difficulty posed by platoons). We’re not going to get that, and failing that, an empirical study like the one presented here is really the only way to improve the model.
Wow, platitudes are coming in hard and fast today.
Turning off winchances is stupid, some people just read them wrong. Especially excessivly high chances inherent the problem of pubbies rushing into battle, trying to “grab some damage”, before its too late. That far too often ends in a disgraceful loss. So when I see a blue’ish winchance, I actually start to play more carefully.
Games below 40% on the other hand are an instaloss if you play normal, so a deep red percentage is like flipping my camping switch on and using the pubbies as cannon fodder.
The one thing this analysis proves – again, is the fact that a skilled MM is indeed possible and Wargaming is just chickening out on it, because they like their game in a state of red where everyone is forced to feel equal. And that’s hardly surprising, since the word “meritocracy” is an offence in the eastern countries.
Oh certainly. One shouldn’t forget that the win-chance doesn’t react to changes in play style due to seeing the win-chance.
I suppose, essentially you should just plays as you would either way – carefully. And just be less disappointed if you lose in a low win-chance game. ;)
I believe what was said at FTR earlies – skilled MM would produce more frustration because nearly everyone’s winrate would go down except the most desperate noobs, bots etc.
The real problem with XVM is when you got a good team but with 5 tomatoes top tank VS a bad team with 5 good top tanks. This match up always goes the same way: XVM says 60% WR, but i know it is false, because our top tier will do shitty things and make us lose ( heavy going in pointless position etc…)
Yeah, had a couple of games where I was in a tier 5 med, and we had 35% chance, then you look at where their good players are, and they’re at tier 3 and you just know it’s a roflstomp coming for you, as well as Ace.
this +1
I dont get why someone would want to see it?
It’s like watchin an average Hollywood movie nowadays, where u know at the beginng how it will end in 80%.
So I rather start a battle, do my best and than I dont care If XVM said we have to win or to loose.
Same reason I dont use playerstats.
But still quite interesting this article. :)
+1
I wouldn’t go so far as to say they mean nothing, but I turned off chance to win a long time ago. I believe the over confidence of a high % can lead to stupid mistakes. I believe this is why you sometimes see low % teams stomping all over the high % team. The flip side is even worse. when some people all but give up right from the start, because of the chance to win, in which case the low % then becomes a self fulfilling prophecy.
Most people are able to look at the team lists and get a general idea of how good or bad their chances are in a particular match up, without the need to put a specific number of it. Personally, I think the chance to win should be removed.
Regardless of precision, it can still be a good indicator. If, for instance, you see that XVM is giving you 27%, then you know to play smarter and possibly more conservatively. The reverse is also true, though I wouldn’t say that if XVM gives you a 95% chance, then go ahead an suicide run. But then, I don’t think most of us needed an article exploring the viability of the statistical analysis behind XVM to tell us that much.
Wrong – look at the winrates and heighten or lower them, the more they further they are away from 50%. In other words: Low XVM win chance means even lower actual win chance, high XVM winrate means even higher actual win chance.
Winchance actually means a lot.
High “winchance” means some retard on your team, who probably has bad stats, claim in allchat: “you are all screwed, we have winchance of *insert number above 80% here*.” And they are going to become mindless lemmings and lose the game.
Low “winchance” means those same retards I described above will go afk and/or drown themselves.
They do mean something. As you can clearly see from these results they mean your chance to win is much higher or much lower than it says but it still lets you know whether a win or loss is more likely correctly.
Does SS have a brain?
If XVM predicts and you win 27% of 40%ers, 50% of 50%ers and 73% of 60%ers then XVM is INCREDIBLY ACCURATE!
If you win 73% of 40%ers and 27% of 60%ers then XVM is inaccurate.
SS, do you need me to explain this to you in more detail?
Nice Info, thanks – and no big surprise to me. Won’t stop people from complaining and ruining the mood before the battle even started though.
Ruining the mood before the battle starts is my favorite part of the game.
siema
Pozz blitz!
You fight enemy tanks or something that was said at start of battle?
Congratulations on taking the time to do the research on this. Well done and thanks for the interesting article.
Time for XVM to update then… at least for those that look at the win chance in the first place. ;)
Yes. It would be interesting to see a response from the XVM guys to this.
Innacruate… oh well
http://en.wikipedia.org/wiki/Normal_distribution
Not necessarily. He took the normal distribution of a sample of 1000 battles with a sample of 29 other, randomly selected players where he was the only constant, and compared it to the normal distribution of 100 of his personal battles. Not the best test for accurate statistical data, but good enough for a personal project, I would think.
So in laymans terms there are three conclusions?
(a) when the win chance is near 50%, ie 51% or 52% it is highly accurate.
(b) when the win chance is 40% and lower it is not accurate and is in fact much worse, similarly where the win chance is 60% or higher it is not accurate, and is in fact much better,
(c) XVM currently uses 1.5 for skill percentage factor, and if it were to use 3.5 it would be more accurate
Is that correct?
yea. you got it.
Nope.
(a) when the win chance is near 50%, ie 51% or 52% it is highly accurate.
–> its is highly (i’d say quite) accurate OVER THE LONG RUN. not battle to battle.
(b) when the win chance is 40% and lower it is not accurate and is in fact much worse, similarly where the win chance is 60% or higher it is not accurate, and is in fact much better,
—> the winchance percentage is not wrong, it is conservative (i.e. leaning towards 50%)
a) of course 50% might be 0% or 100% for each battle. but isnt this the purpose of winchance? to determine the result over more battles?
b) can you extend this? and without making 1+1=2 <=> ln(e) + sin^2(x)+ cos^2(x) = 2. remember we are in layman’s area here.
edit: cant find a way to write the equivalent sign…
=
Exactly. % is 100, not 1 as some would like to think.
Actually 1 % is 1/100 = 0,01 and therefore 100 % is 100/100 = 1.
“You win for example 27% of 40%ers, 50% of 50%ers and 73% of 60%ers.”
maybe xvm scales percentages like that so it won’t demoralize by showing the real 27% chance and it won’t make you overconfident on 73%ers.
also study is biased. people act different on low chance to win pressure. some tend to derp thinking they have no chance, others tend to play carefully and at least farm some dmg if not win the game.
You can increase the accuracy of the prediction, by basing it on the actual skill of both teams. A 70% chance with tomatoes is utterly unreliable, in fact, it might cause over confident tomatoes.
Also to remind the huge statistical study of nesseria (also more than 1000 battles) : http://forum.worldoftanks.eu/index.php?/topic/379636-analyse-objective-du-mm-via-xvm-sur-plus-de-1000-batailles/
At this point, nesseria found on his side a good accuracy of the Win chance in XVM (see his figure in part III – Results)
And he reminds that the CTW value depends of the player himself !
Nice work, I did a similar thing on a set of 5000 battles and my conclusion was that the spreads were way to huge to do something about it formula wise (of course it can be improved).
I had a nice average WC of 43% but a WR of ~52% with an error of 5% at FWHM. In other words you might be able to correct for the inherent offset between any old and new WC vs. WR prediction but how are you going to handle the spread in WR for one WC?
One thing would be for WG to balance their battles better but when that doesn’t happen that spread will remain there.
Guessing it would be bettter if tier of the players wold taken in to considerartion. If the top tier heavies on your team are total tomatoes, a couple of low tier players willl have a hard time compensating for that
I think observation is dependent on author’s individual skill to win battles and tanks he plays. There are tanks with marginal/random/map dependent impact on battle as arty or light tanks, there are tanks that can consistently influence battle to great extent when played by really good players – as high tier meds, or even above average players – pre-nerf Hellcat, T67.
XVM only gives an accurate specific number between 45% and 55%. From that point on, it is screwed.
I mean, you lose 95% of the matches with less than 40%wc and you win 95% of the matches with more than 60%
Bearing this in mind, it is a great prediction tool.
If nobody had XVM it would be accurate as all hell since idiots wouldn’t hit quit button the moment they saw they had ‘only’ a 40% chance to win.
How would you want to have same win chance in avg. as your win rate….. That’s bullshit, there are 15 players on the team. You are evaluating with 1 variable. Also he didn’t mention platoons, which can affect WC and WR a lot.
Doesnt it make sense? If you play 1000 matches, in the long run the expected win chance would be the same as YOUR win chance because you are the only stable factor in your games.
Great analysis, so nice to see someone doing a little bit of math :) Just label the graphs better next time. So let’s have that implemented in XVM. Boom, done, problem solved.
If i understood correctly the actual formula
P(win) = ((Ka / (Ka + Ke) – 0.5) * 1.5 + 0.5) * 100%
should be corrected as
Pnew(win) = ((Ka / (Ka + Ke) – 0.5) * 3.5 + 0.5) * 100%
Why not submit to xvm team the correction? or is player dependend?
so do you want the win chances to add up to 200% (think before you post). Please.
I understood that, with 3.5 coeff, the prediction chances will be accurate.
[ Quoting: ...Basically it scales the skill percentage with an arbitrary factor of 1.5. According to the figure above, the scaling factor should in fact be 2.3 times higher, 3.5, to produce more accurate results... ]
Isn’t so?
It will only be accurate for a player with a skill level similar to the author’s.
Also, If you play much better than in your previous games (via learning, improving your crews, tank moduls, adding equipment, shooting gold when you previously wouldn’t have), but your XVM rating is based on the average of all games, the predictions will be inexact.
Therefore you could use a version of XVM, where the win-% is bases on the last 1000 battles of each player of both teams, so it would be much more precise.
All this effort on the skill component of the formula ignores the real problem in XVM’s chance to win – there is no weighting of the relative efficiency of a particular tank, only the tier of the tank is taken into account. For the purposes of the formula, an AMX40 = Hetzer = SU-5. This can cause some very lopsided inaccuracies in the prediction. One would assume those to be random and balance out in large sample sets, but the reality is that server-wide, some tanks are preferred over others and are more likely to show up in any given match, which will cause studies based on only one player to skew depending on that player’s fleet of tanks. To really get at the accuracy of the formula, pull a large sample set of players who own every tank available and run a large-sample randomized test.
XVM is a reasonable guide to how most teams will perform. When your win chance is between 40-60%, it can go either way dependent on how people perform, and RNG.
However, when you get a win chance of < 30% (which happens far too often), you will lose the majority of those games (you can win the odd one…).
More often than not, the problem isn’t whether XVM is accurate, the problem is people using XVM convince themselves of how the battle will play out.
Too many people see a low win chance and simply give up because they believe it’s a loss from the start, so they YOLO down the middle and die instantly.
On the other side, people in a team with 80% win chance think it’s an auto-win, so they YOLO down the middle thinking they’re going to kill everything in their path, and die instantly.
XVM doesn’t even take into account how good a specific player is on a specific map or in a specific vehicle. Of course it’s wrong alot.
1k games is a ridiculously small sample size, you’d know that if you did any actual research in real life.Like, you’d get laughed out of the room.From personal experience, xvm seems to be about 60-70% accurate and that’s over 3k games.
for ne win chance gives an opportunity to take a look at my team’s capabilities. And you should first take a look at your team’s top tanks – if they are red/yellow/orange your win chance is lower two times than it iis shown by XVM.
Well, conclusion that win chance is inaccurate is kinda obvious, weather forecasts are also inaccurate due to the amount of variables. You’d have to use neural networks processing huge amounts of data adapting to the gameplay changes, buffs/nerfs, different game styles etc. to get satisfying-ish predictions and I think it’s just not worth the effort.
For example I got rid off my XVM because I wasn’t sure if the additional informations are helping me or crippling me :)
Of course xvm prediction is bullshit, I’ve never won battle with over 80% win chance, when people see it, they all get some kind of monkey brain and mostly die in 4 minutes.
I’ve seen that quite a few times, and it’s annoying as fuck -
“80% chance of win? We’re invincible!!1!!11!!!”…
Ftr reached new level of bottoms to reach…. me sad
Is this change to win prediction still implemented in XVM? I couldn’t find the option to enable it lately.
Still working for me.
Very interesting
Pingback: Jak dokładne są przewidywania XVM?
Lol so u´re spending resources on testing the most obvious things. Next pls do research on “Are players with highest WN8 those with highest dmg per battle?” or “how often does a player with 60%WR win ?” or u know what, do something valuable and get into testing the depths of RNG
XVM WC is complete bullshit…
WC should be left in the WC….
Well win chance must not really taken into to acount cause it wot it’s randoms sooo everything can be possible (lose a game with full unicums in your team/get annihilated in the first 5 minutes and so on) best thing is that xvm on it’s active services have an option to not display the win chance and it’s the best option :P
you can understand when a game is going to be a loss when you are 5 minutes in and you lose 7 tanks already
win chance even if they update the formula cannot preddict accurate enough what will happen inside the game
it will give you an idea nothing else
Basically, on low w/r chances, you have a team based of mostly bad players, on which you can’t rely on doing “this or that” rationaly. There is a higher anomaly factor with bad players, meaning (on a large battle number scale) that they could only do worse on the battlefield, not better.
On the other hand, higher w/r chance means you are stuck with mostly good players, there is a high probability that they know what they are doing, and anomaly factor in the battle is smaller. Also, good player (on a large battle number scale) reacts faster to the changes on the battlefield, and has a greater chance on turning the battle in your odds.
In conclusion : There are few GOOD players in the game, but there is sh*tload of BAD ones. Distribution is not linear.
Hmm odd, someone on the US forum made an experiment like this ages ago. He had more battles in the experiment and his experiment showed to be accurate. Maybe some tiers are more accurate than others. Maybe back then tere were less rerollers to scew the %.
http://forum.worldoftanks.com/index.php?/topic/174317-1000-games-with-xvm/#topmost
The experiment was from 8.5 though and back then less people rerolled.
ADD TIER WEIGHT and all will be more accurate. like a unicom who is a top of list is not the same as if he is in the tail of the list (Except light tanks of course) they must add non-linear variant that represents a tier strength (not 10 for tier X and 3 for tier III).
I use this method in my head and it’s >95% accurate.
Good luck
“Win Chance” is ruining the game. Who is the retard who thought this was a good idea? Every god damn game I have to hear from someone in my clan, we only have a 28% chance to win might as well give up. Just another mark on the wall towards this games downfall.
Test rows miss the obvious: What tank was he in. Some have more effect on the battle than others.
Just don’t use the win chance, duh. Use XVM if you want, then just use it to see which enemy that you should be very careful of and see which ally that you could trust at some degree (or not at all). Then carry the game as usual.
i turned off win chance in xvm, by seeing the tanks+skill of your enemies vs your team tells you just enough of the battle result
If win chance was reported accurately then there would be more players bailing on low chance games and more players going AFK in high win chance games. In my view the XVM win chance is loaded to prevent breaking the game.
Using XVM stats but without this stupid WinChance…this isn’t really motivating and actually one of the worst features of XVM.
Interesting article, though it might be prudent to point out that his samples are somewhat flawed, as well as the fact that statistical theorems tell us that as the sample size of the number of battles increases, the better his results may fit the XVM rating. However, on a battle-to-battle level, I don’t think we need an accurate statistical study to see that XVM isn’t the best indicator of a win or loss. That said, XVM can’t tell you if you’ll win or lose, but it can tell you how to play. For example, if it’s giving you a 27% chance to win, then you know to play smarter and more conservatively than you usually would. Alternately, with a 95% chance, you know you can take some larger risks and rely on your team to cover you or support you.
Pingback: [WoT PC] Verschiedene Änderungen auf dem Supertest | Wargames-News
Isn’t it possible to calculate the variance of the normal distribution every battle accounting for every other player win ratio variance?
First you calculate the win ratio of every player, then its variance, then the battle chances to win, then the variance of the latter.
If you are experts in the matter let me know.
I’ve always wondered the same thing, so I kept stats for a while (and still do).
For 7652 battles I have played, I recorded on a spreadsheet the winning chance that XVM gave, and whether it ended up being right or wrong (I’ve played about 1000 more matches but for various reasons couldn’t get data: XVM server down, WoT crashes, etc.)
Now, I have my spreadsheet set up so that any prediction that falls in the 46% to 55% range are simply ignored. The reason for that is that I believe any prediction falling in that middle 10% bracket (50% ± 5%) would essentially be too close for XVM to call accurately: the fight could go either way for reasons totally independent of player skill, like lag, the MM stacking 8 vs 3 heavies or the never-to-be-sufficiently-damned RNG.
Results: Total sample size: 7652 matches
Middle bracket: 2493 matches with a 46%-55% chance to win. That’s 32.6% of all the matches, roughly one in three that you can call “even odds”.
Significant bracket: 5159 matches (67.4% of the sample size) had a predicted chance to win below 46% or over 55%.
Prediction accuracy for the significant bracket? 72.4%. Meaning that for matches that have a significant chance to be lost or won by your team, XVM has accurately predicted the outcome a little more than 7 times out of 10.
I’d say that overall, it works fine.
Why is it wrong almost 30% of the time? In part, because it’s statistics, not magic. There will always be dots outside the curve. And in part because of player psychology. XVM assumes that every player will play to its exact potential as expressed by WN8, whatever the predicted result. Sometimes players see a 75% chance to win and decide to sit on their fat asses expecting victory without making any effort: they defeat the prediction and earn their pie in the face. Sometimes players sacrifice all common sense to the completion of a personal mission and give the victory to the 25% team. And sometimes, the RNG is simply a fucking bitch and on the other team’s side!
Does SS have a brain?
If XVM predicts and you win 27% of 40%ers, 50% of 50%ers and 73% of 60%ers then XVM is INCREDIBLY ACCURATE!
If you win 73% of 40%ers and 27% of 60%ers then XVM is inaccurate.
SS, do you need me to explain this to you in more detail?
All these graphs are not very useful if the statistician doesn’t provide us with data that is useful in practical application. Finding the apparent relation between XVM winrate and actual winrate is useful, but it doesn’t tell us much in practice.
This is why we need Bayesian inference. For this, I’m going to find something possibly more useful in a practical sense: probability of winning given a good XVM win rate prediction.
For this, I’m going to classify a good win rate as XVM showing >50% chance of winning.
From the graph we can tell:
Games Won With Good Win Rate:
20 + 28 + 25 + 32 + 34 + 30 + 29 + 25 + 32 + 20 + 20 + 28 + 24 + 23 + 22 + 21 + 12 + 9 + 7 + 9 + 6 + 8 + 6 + 2 + 1 + 1 + 4 + 2 + 4 + 3 + 2 + 1 + 1 + 1 + 2
= 494
Overall Win Rate:
Spitfeuer117 has 55.90% win rate
Games with Good Win Rate (I’m going to go for >50%):
55 + 39 + 50 + 59 + 43 + 42 + 34 + 51 + 28 + 27 + 26 + 31 + 27 + 27 + 26 + 17 + 10 + 10 + 10 + 6 + 9 + 5 + 2 + 1 + 1 + 4 + 2 + 4 + 4 + 2 + 1 + 1 + 1 + 2
= 657
Total games: 1000
Given:
GW = % games won with good winrate = 494 / 1000 = 0.494 = 49.40%
W = overall win rate = 55.90%
G = % good win rate = 657 / 1000 = 0.657 = 65.70%
P = % probability of winning given a good winrate
GW * (G/W) = P
0.494 * (0.657 / 0.559 ) = 0.5806 = 58.1%
Thus, for this specific player, probability of winning given a good win rate is 58.1%.
A possibly more interesting probability would be the chance of loosing given a bad win rate.
I will say my maths is far from perfect, so do check if you don’t trust me.
tl;dr : I calculate Spitfeuer117 has 58.1% chance of winning games with more than 50% XVM predicted win rate.
Out of curiosity, I did the converse:
Games Lost with Bad Win Rate: 1000 – 494 = 506
Overall Loose Rate: 100% – 55.90% = 44.1% (only if ¬win = loose)
Games with Bad Win Rate: 1000 – 657 = 343
Given:
BL = % games won with good winrate = 506/1000 = 0.506
L = overall win rate = 44.1% = 0.441
B = % good win rate = 343/1000 = 0.343
P’ = % probability of loosing given bad win rate
BL * (B/L) = P’
0.506 * (0.343 / 0.441 ) = 0.393556 = 39.4%
Chances of loosing a bad winrate game = 39.4%
Laifs,
All I had to do is look at Spitfeuer117′s statistics on Noobmeter to see that on a sample size of 20158 battles, he won 55.90% of them. Of his last 1000 matches, he won 58.05%. For the sample he selected, 59.8%
My point is, you didn’t need XVM to find that HIS chance of winning a random match is about 58%: that’s true irrespective of any WC XVM predicts for a particular match.
The only problem I have with his approach is that he limited the data sample to those matches that fell in the 44-66% prediction range (those for which there was 20 or more matches played). Why? He discarded perfectly valid results for roughly 170 matches.
Look (on xvm1.jpg) at all those Predicted-WC-below-44% matches that were lost (as predicted) and those Predicted-WC-above-66% matches than were won (again, as predicted). Adding those to the “prediction vs result” graph (xvm2.jpg) would have extended the number of dots in both directions and could have made a difference. I suspect the “slope” or scaling error might have actually gone closer to the x=y accuracy goal.
I use a simple yet effective formula. I count the number on ‘Siemka” written in my team’s chat.
1 Siemka = We still have hope
2 Siemka = Little hope
3 Siemka = All hope is lost
PS: Sorry polish people for this joke, didn’t mean to be rude or sound racist.
an equation that doesn’t use Tiers in it is POINTLESS.
it’s just plain stupid !!!
XVM prediction isn’t accurate all the time BUT:
- If you see that you have 40% chance to win, but you also see that is because of some bottom tier platoon, you can read those 40% as maybe 50%.
- You have to watch stats of more dangerous tanks. If you see that their top tiers are read, and bottom tiers are blue, and your team has better top tiers, your chances are a bit higher, right? No point if someone drives a good tank if he doesn’t know how to use it…