Would it be possible from the data you have at hand to make out some comparison, like whats the average score, winrate and K/D, for example on rank III vehicle or vehicles, for both arcade and realistic?
An 80% player is a lot better than top fifth, probably closer to top 0.1% of players.
The 80% player would be at the very end of the graph where only a tiny population is.
Ordinarily yes, but this is ordinal data, not quantitative.
“80%” here means “my average score and game position is greater than 80% of all the other players in matches with me (and less than the other 20%).”
There’s definitely still going to be some belling (because as the number approaches 100 it gets harder and harder to keep up) but it’s harder for me to say what the shape of that curve would be.
Just looking at the distribution of points on the graphs also suggests a greater concentration of points in the middle and less at the ends. Larger player samples would give a better idea what the distribution is. I’m the sample I used there were only 2 dots above 80 in ground AB and none in ground RB.
1 Like
Working on something like that at the moment actually, when the grind permits…
In the meantime, another interesting inference one could draw from the data is, if people accept the lines on the scatterplots are a reasonable first guess (yes, more data would be better, granted), then plotting your own dot on the graphs provided (from your ARP and score on your service record) says something about the natural efficiency of your play as opposed to what it should be for events. It helps that a lot of this last month was also an event, in this regard, so if you played it that could be useful to know.
The biggest factor in that, especially if you’re below the line, is going to be BR, but if you’re above there’s likely also an element of doing things that allow you to really run up absurd scores in those few games when it all comes together for you too.
In the sample used there wasn’t a lot of up and down variation for the ground modes, suggesting there aren’t a lot of killer BRs in ground with those kinds of “natural” efficiencies. There’s still the artificial disincentivization Gaijin is applying against playing lower ranks, though.
In air that’s different: a person who does the “Ar-2 arbitrage” strat for events will likely place well above the line and have an easier time of it (as one example).
And if you’re under that line and planning to do an event in that mode, yeah, either switch BRs or identify those “run the table” kinds of strats. Note this is separate from strats that just make you a better player generally, which would just move you up and down the line, not vertically offset from it. If you’re above the line by a lot you’re doing something right that only clicks in when you’re at the top of the team (or as I suspect with naval, but also air to an extent, there’s actually a couple of performance lines in the data for the different BR ranges, not just one, and you’re on the “good” line).
2 Likes
Could you please mention me if you put it together? I would be interested in the results.
The data you used to get the average scores was from normal gameplay or from playing the event? I’m just asking cause I took the time to calculate the average score on my matches while doing the 3rd mark and I’m getting a slightly higher average score than you but that could be explained for the fact that during events some players play “harder” or optimize score.
I got an average score for Ground AB of 1190.
I’d say there is little point in using normal gameplay data if you want to estimate event duration. I most certainly play score optimized when doing events, unless I don’t need to worry about time. And I don’t think I am alone with this.
1 Like
So your service record gives you your average score and average relative position by mode for the last 30 days (as well as lifetime), if you’ve played 100 games or more in that mode. That’s all I was using for this sample of players. As far as we know, it’s good data, although the amount of lag on it is unclear.
People are going to optimize, sure. In this case, “last 30 days” includes a lot of the previous event, so that helps. There are also limits on how much optimizing you can actually achieve when everyone else is also optimizing against you. The flip side of this is the 1200 point player saying they can “turn it on” and become a 2000 point player for events if they feel like it. My experience is that, generally, they can’t and they have a very inaccurate idea how long this event will take them, and others.
The graphs actually show this for ground. If you could just “turn it on and off” statistically there would be more vertical variation. You could make that case here on the data for air and naval, I think. But if all the dots are all in a line, as it is with ground with this sample, that means score and ARP are more tightly tied and all you can do to get more RAW score per hour faster, really is:
A) reduce average match time by quitting faster, as has been suggested already, or
B) do better than you do on average position in your team than you do normally.
Which, again, is going to be non-trivial if the rest of your team is doing the same thing as you. If you’re a top-third player, and you go to a lower BR and an optimized lineup, you still have to actually BECOME a top-quartile player by doing so to actually see event performance gains.
Your point also raises the question, now that the format has changed and there’s always an event, if there’s ever any non-optimized time again.
2 Likes
Yes, I believe in you but what I was saying is that players usually optimize playing during events, just like Dodo_Dud also mentioned, so it’s likely that the average scores will be a little higher than with normal gameplay.
The average score I got during the event is slightly higher than the one in your table which could be explained by just that. But the value is pretty close.
I mean if you’re generally the only player on your team who cares about the event (which could be more likely in AB than RB), then your driving hard to the hoop is going to tend to improve your relative position. If everyone else is also trying to finish the event early, it’ll tend to cancel out.
What we don’t really have a good sense of is how much of the player base gives a toss about any of this. Forum participants are going to be more invested in this stuff than the average player, I suspect, but we don’t know by how much, so we can’t really account for it as a variable.
1 Like
That’s not how a gaussian works and they are talking about percentiles anyways. An 80% player is within 1 sigma as the ± 1 sigma band contains approximately 68% of the population (0 sigma would be the median player, i.e. a 50% player). So a player at exactly +1 sigma would be an 84% player. Top 0.1% is about 4.5 sigma.
Updated the post since you guys were asking:
UPDATE: Okay, some people have brought up what percentage of people we’re talking about in each of the categories above. We can crunch the sample we have (N=208) through a standard deviation calculator, assuming the distribution is normal for these purposes and the sample representative, and see what percent of players are better or worse than a given average relative position on their service record. The standard deviation on the sample is 12.502, on a mean of 51.827, if you want to follow along at home. What it comes to is:
50% player (by average relative position) in the sample: Better than 44.2% of all players.
67% player: Better than 88.8% of all players
75% player: Better than 96.8% of all players
80% player: Better than 98.8% of all players.
You can run the same math in reverse, too. Assuming all the many many assumptions above, for an average score of 1500 in ground RB, which is equivalent to the 61% percentile of game results, you can infer that 76.7% of ground RB players will not be able to achieve that score.
2 Likes
For future reference, here’s a breakdown of average relative position values (%ARP) and population of active gamers who can equal or beat them (%POP), assuming representative sample, normal distribution, and sigma and mean as given above:
| %ARP | %POP |
|---|---|
| 50 | 55.81 |
| 55 | 39.98 |
| 60 | 25.66 |
| 65 | 14.60 |
| 70 | 7.30 |
| 75 | 3.19 |
| 80 | 1.21 |
| 85 | 0.40 |
| 90 | 0.11 |
| %POP | %ARP |
|---|---|
| 50 | 51.8 |
| 45 | 53.4 |
| 40 | 55.0 |
| 35 | 56.6 |
| 30 | 58.4 |
| 25 | 60.3 |
| 20 | 62.3 |
| 15 | 64.8 |
| 10 | 67.8 |
| 5 | 72.4 |
| 1 | 80.9 |
| 0.1 | 90.5 |
3 Likes
This is a very good post but it cannot be, i cannot possibly be better than 98.8% of all players in ground arcade battles. I can do 40k mission score in 2.30h MAXIMUM if i play rankVII, even less if i play something broken like the italian sherman lineup at rank3 (7k avg.mission score per battle, 6 battles per hour).
Going of your stats, yes ,yes you are XD
May i ask what your average team placement is for arcade ground?
Here’s one thing the OP did not consider:
- Rank bonus.
OP considers only 1.0 rank bonus and only mode based multipliers.
Dunno ground multipliers, but IIRC for air we had…
Rank 1-4: 0.9
Rank V: 1.0
Rank VI: 1.1
Rank VII. 1.2
Rank VIII: Can’t remember.
The new profile cards have made it much harder than it was to check other players stats, sadly. I could never have scraped this same data today. But I think the math holds up, it’s certainly never been challenged.
But IF your ground AB average score was 7k per game, another way to look at it is these numbers suggest a standard 16 player side has about 20k score on average to split between them in an average AB game. If you’re taking a third of it on average, that leaves 2/3 for the other 15 players, so to be on a team with you they’d only be making on average about 4-500 score per game. Do the math on how long it will take them to do this event under those conditions… It’s not pretty. I don’t think the players on either team would much enjoy that kind of results differential.
There, I’ve proved mathematically that seal clubbing is bad. :)
No way to consider it, starting from the initial data set. But given that it’s likely offset in many modes by smaller match sizes and more ODLing at higher tiers (reducing total available score) I suspect the impact isn’t much. If anything it might not offset those factors enough. The bigger effect will be if you slide yourself up or down that score-per-game line in the original graphs by playing at a level either too easy or too hard for your skill level.
Sorry for the late response, my average team placement is 80% in 13674 battles, but take in consideration that i also play navalAB in which i’m really bad so i rarely place top3…
Oh yeah, you’re absolutely better than almost all players. Are you familiar with a normal distribution bell curve?
If you are, then imagine a player with an average placement of 50% and think of how their curve will look, and now imagine how yours at 80% will look.
To have an average of 80% half your games will fall above that and half will fall below (on average). BUT Since you cannot get better than first (100%) then if you get one game lower than 60% you have to compensate for that by getting more than one game above 80%. So say you get 40% in one game you have to get two games at 100% to compensate for the change in average. So the amount of games you have above 80% placement has to be drastically more than the number of games below to achieve the average of 80%.
Just one game where you are last (0%) has to be compensated with 4 games att 100% to get an average of 80% (400/5=80). So you can imagine how often you have to get first or second in the team to maintain an average of 80% :)
I hope I explained it well enough :)
1 Like