Source: wot-news.com Hello everyone, today, we have something interesting, an analysis of the new personal rating by Edrard. For those who do not know, Edrard is the creator of Efficiency Rating and the owner of wot-news.com – so he is highly competent to comment on this latest Wargaming feature. Edrard: And so it was that with the patch 8.8, Wargaming decided to join the list of player rankings with its new invention, called Personal Player Rating (furthermore as PR). This article will consist of a small analysis of this rating, will mention good and bad sides, but whether to use the PR or not, that’s up to you. So for starters, let’s have a look at the formula: bc – amount of battles played (“battlecount”) win – winrate (interval 0-1) surv – survival rate (interval 0-1) hit – hitrate (0-1) xp – average XP per battle dmg – average dmg dealt per battle The formula seems complicated at first glance, but it is not – its parameters can be seen clearly, including the modifiers and their influence on other values. For this article, I will use a (random) player with following parameters: bс = 4000 bc088 = 100 win = 0.49 surv = 0.5 hit = 0.7 xp = 400 dmg = 1510 At first, let’s have a look at winrate influence on your rating: f(x)=(2/(1+exp(-(4000)/4500))-1)*(3000/(1+exp((0.5-x)/0.03)) + 7000*0.3 + 6000*0.25 +5*(2/(1+exp(100/500))-1)*240 + 1340) win rat 0 2011.6566893 0.1 2011.6586446 0.2 2011.7134535 0.3 2013.2478847 0.4 2054.7807946 0.5 2637.639092 0.6 3220.4973894 0.7 3262.0302993 0.8 3263.5647306 0.9 3263.6195394 1 3263.6214948 Here we can see that the influence of one percent winrate (regardless of how high his winrate is) is principially small. I personally never liked the idea of using winrate for one big reason: it’s not a definitive parameter – you can be an average player playing in a good platoon or company and have high winrate while not contributing anything to it. I suspect that in next patch, the statistics will be separated, at least by removing the companies and clanwars from it (just guessing), but the platoons are not going to be separated and I myself know of average players, who – despite not being skilled – have 55 percent winrate because of a good platoon. It looks like playing in platoons is beneficial, if you bring friends to the game, some of who can be good players – it will improve your rating. Or on the contrary – you can be a good player, who has bad luck of getting dropped into noob teams and therefore never improving the winrate over 50 pecent. Next there is the survival rate, the connection between rating and survival rate is linear and not interesting in particular, for our player it will be like this: f(x)=(2/(1+exp(-(4000)/4500))-1)*(3000/(1+exp((0.01)/0.03)) + 7000*(x-0.2) + 6000*0.25 +5*(2/(1+exp(100/500))-1)*240 + 1340) surv f(x) 0 1073.6383124 0.1 1365,7634674 0.2 1657,8886225 0.3 1950.0137775 0.4 2242.1389326 0.5 2534.2640876 0.6 2826.3892426 0.7 3118.5143977 0.8 3410.6395527 0.9 3702.7647078 1 3994.8898628 This means that for one percent survival rate, you recieve roughly a 2,13 percent better rating. This parameter in general represents the player well, I wanted to introduce it to the Efficiency Rating for ages and I will definitely introduce it to Efficiency Rating 2, if there ever is such a thing. On the other hand, it might make people care too much about their survival. I don’t consider “ololol rush” to be a good strategy, but to sit in a bush when your team is attacking is not good either. Next is the hitrate, here again the corellation is linear: f(x)=(2/(1+exp(-(4000)/4500))-1)*(3000/(1+exp((0.01)/0.03)) + 7000*(0.3) + 6000*(x-0.45) +5*(2/(1+exp(100/500))-1)*240 + 1340) x f(x) 0 781.5131573 0.1 1031.9061474 0.2 1282.2991374 0.3 1532.6921275 0.4 1783.0851175 0.5 2033.4781075 0.6 2283.8710975 0.7 2534.2640876 0.8 2784.6570776 0.9 3035,0500677 1 3285,4430577 Its influence is such that for every hitrate percent more, the rating improves by 0,8 percent. Let’s say it like this: this parameter also is not definitive, it is not clear what it represents in the formula, the skill to hit a silhouette, or possibly the favor of the RNG? I would understand if the percentage of shells that actually did damage was taken into consideration, but considering the fact the formula has already damage as a parameter, that would be redundant. Next is the XP, tied to battles played, but that will be shown later, let’s skip it for now and deal with the average damage: f(x)=(2/(1+exp(-(4000)/4500))-1)*(3000/(1+exp((0.01)/0.03)) + 7000*(0.3) + 6000*(0.25) +5*(2/(1+exp(100/500))-1)*240 + (x-170)) x f(x) 0 1904.108396 100 1945.840561 200 1987.572726 300 2029.304891 400 2071.037056 500 2112.769221 600 2154.501386 700 2196.233551 800 2237.9657161 900 2279.6978811 1000 2321.4300461 1100 2363.1622111 1200 2404.8943761 1300 2446.6265411 1400 2488.3587061 1500 2530.0908711 1600 2571.8230361 1700 2613.5552011 1800 2655.2873661 1900 2697.0195311 2000 2738.7516961 2100 2780.4838611 2200 2822.2160261 2300 2863.9481911 2400 2905.6803561 2500 2947.4125212 2600 2989.1446862 2700 3030,8768512 2800 3072,6090162 2900 3114.3411812 3000 3156.0733462 It is so that 100 average damage more improves your rating by 1,37 percent. To be honest, it’s strange to see such a small influence of average damage on PR, I always considered it one of the main parameters and it is good that it influences the PR at all. It’s difficult to say how big or small its influence is, for that you’d have to somehow rate the difficulty of improving your hitrate by 1 percent and damage per battle by 100, but subjectively I think its influence is too small. And now the “fun stuff”: there are two parameters left – a multiplier, incorporating the average XP per battle in patch 0.8.8 and the amount of battles. I didn’t want to create three-dimensional graph, it also makes no sense to make a graph of the corellation between the average XP and rating, I will only say that the influence of average XP per battle on rating will change based on how many battles did the player play from 0 to 4 percent. Average experience is a very important parameter and the fact that Wargaming decided to use it without the premium account bonus is correct. If such a separation was implemented into stats in general, noone would have to invent any ratings and people would only look at the average experience, but unfortunately that’s not possible anymore. The influence of the average XP is completely sufficient and the fact it is dependent on battles played is correct, but unfortunately the average tech tier parameter was not used so that the use of low and high tier tech was balanced, but it’s not all bad (apart from the amount of battles needed to reach maximum rating). Let’s have a look what will happen, when the amount of average XP doesn’t increase, only the battles played Continue reading →
More...
Bookmarks