paper: Win Rate Data

It’s probably more of a feedback loop, realistically. Teams who play more get better at it, and teams that are good are more interested in playing more. It makes sense that it would have positive feedback.

2 Likes

The three teams closest to 600 matches and 30% win rate are (without identifying them):

481 matches, 37.2%
617 matches, 44.7%
684 matches, 40.9%

There is something other than winning matches driving the three teams identified above. If you were able to get data that could measure whatever that is, I suspect these three teams would be at the top of the pack rather than the bottom.


The data is probably driven both by teams that win more tending to want to play more and by teams winning more as they gain more experience. I don’t think this is a simple independent/dependent variable situation, there’s probably a good number of confounding variables.

Everything here pretty much makes sense. Win more, play more. Also not surprising to see powerhouse regional teams to the far right, and older powerhouse Michigan teams to the top. It’ll be interesting to see if teams like 2056, 1114, 118, and 148 move closer to that 70% mark where alot of the good Michigan teams are, as they spend more time in the district model. It would be cool to see a few of these graphs but broken down by regions and team age.

2 Likes

The win rate is cumulative along the bottom. The line of zeros growing on the left is because I couldn’t find an easy way to get rid of the teams who weren’t competing anymore. They can just be ignored.

4 Likes

I’d be interested to see this graph+trend line for all active FRC teams (rather than starting at 600+ matches played) :face_with_monocle:

1 Like

Here’s a plot of lifetime win rate vs team number, which is a (somewhat bad) proxy for team age. Interestingly, the trend we saw for number of matches played all but disappears. There’s only a slight improvement in mean win rate with age it seems.

[edit] There’s slightly more of a trend for teams that managed to last until 2019:

2 Likes

Jagged race to the top.

2 Likes

I was digging through some of this data a bit more, I found that it says 4613 and 4976 had no matches played in 2019.

I’m not sure how prevalent this issue is, or if it affects years other than 2019, but I thought I should let know something is a little off.

1 Like

I wanted to look at this a bit more directly:

For every team with a rookie year of 2004 or later, I grabbed the winrate for their first year, second year, third year, etc.

This is plotted below in a histogram of 5% win rate bins, split over 16 years. Each year of data is normalized to itself, as there are 5522 teams who had a first year since 2004, but only 83 who had a sixteenth year since 2004.

It’s a busy image, but the conclusion seems to be that teams with more experience tend to have higher win rates. This can’t tell you the strength of that effect, but it is visually noticeable.

Of course, this doesn’t necessarily mean that teams are getting better with age, the teams with lower win rates could just be dropping out… Stay tuned for more…


Here's the plotted data in case anyone wants to play with it
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
0-5% 48 139 104 78 68 46 42 26 13 17 9 3 1 2 0 0
5-10% 15 15 13 6 2 2 3 1 3 0 0 0 1 0 0 0
10-15% 169 103 102 59 38 40 21 21 15 8 10 3 5 2 0 0
15-20% 104 70 47 27 22 17 13 16 10 7 9 8 4 2 1 1
20-25% 258 206 155 102 87 67 65 39 37 28 23 17 9 5 1 1
25-30% 441 349 254 205 145 99 104 56 57 32 39 27 13 10 3 0
30-35% 708 496 364 296 254 174 182 121 84 77 65 46 42 12 9 5
35-40% 538 419 306 248 224 184 151 128 104 72 60 30 33 18 10 4
40-45% 872 646 543 461 364 313 224 220 151 134 91 73 55 33 27 12
45-50% 420 383 299 275 228 190 171 129 129 88 71 62 39 32 14 8
50-55% 879 715 563 485 382 344 267 242 189 157 121 99 51 50 24 14
55-60% 446 352 292 269 267 213 177 173 117 98 81 80 33 31 17 12
60-65% 349 287 246 218 178 183 134 117 100 86 57 42 33 23 11 7
65-70% 134 129 112 112 110 95 74 63 55 46 36 25 27 20 10 8
70-75% 94 63 61 64 41 51 36 45 31 18 20 20 18 7 0 3
75-80% 39 35 27 33 33 29 25 26 17 12 16 10 7 7 4 4
80-85% 6 16 19 7 14 5 9 10 4 9 6 3 5 1 4 3
85-90% 2 6 8 1 4 3 2 4 5 3 4 2 2 2 0 0
90-95% 0 1 1 1 0 1 2 0 2 2 2 0 1 0 1 1
95-100% 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
5522 4430 3516 2947 2461 2056 1702 1437 1123 894 720 550 379 257 136 83

The graph showing teams existing in 2019 appears to have a strong trend line, starting with a mean around 40% for rookies, and reaching beyond mid 50% for sub 1000 teams.

This is an interesting depiction. There’s a clear trend break at 45%–teams appear to take divergent paths toward either mediocrity, moving into the 2nd largest bin or 40-45%, or moving steadily ahead in winning more as the program matures. There also seems to be a distinct break at about 11-12 years when a set of teams make a jump to a 65% win rate and beyond.