Posted by schwarze on 8/9/2020 11:37:00 PM (view original):

Posted by brianjw on 8/9/2020 10:09:00 PM (view original):
Do you have the advanced standings at the all star breaks?

It would be interesting to compare results in the second half with (a) results in the first half, (b) exp win % in the first half and (c) record in 2+ run games in the first half. When I did that one time (b) was the best correlate, which was surprising because I really expected it to be (c).

Which of these best predicts wins, or best predicts the others, is arguably the one that is best indicator of skill rather than luck. I still have a gut feeling that it's record in 2+ run games, but it would be good to back that up.

Sorry - I never got the advanced standings at the half-way point. (I did ask for ozomatli for them).

I think I can reverse engineer the splits from the spreadsheets once WIS publishes them. If I get some time later this week I might have a look at what first half stats best predict the second half.

Posted by brianjw on 8/9/2020 10:09:00 PM (view original):
Do you have the advanced standings at the all star breaks?

It would be interesting to compare results in the second half with (a) results in the first half, (b) exp win % in the first half and (c) record in 2+ run games in the first half. When I did that one time (b) was the best correlate, which was surprising because I really expected it to be (c).

Which of these best predicts wins, or best predicts the others, is arguably the one that is best indicator of skill rather than luck. I still have a gut feeling that it's record in 2+ run games, but it would be good to back that up.

This really has to be dependent on the huge mop-up blowouts. The fewer of those there are, the better exp win% is.

There's still both (a) a fair amount of luck even in 2+ run games and (b) a mild amount of skill in 1 run games, so there is a hurdle to clear to beat exp win%.

Posted by Jtpsops on 8/9/2020 11:18:00 PM (view original):
I'm not sure what it means, but my actual (.545), expected (.542), 1-run (.548), and 2+ run (.544) winning percentages were all unbelievably close.

Does that mean I had pretty much no luck, good or bad?

I don't think we should expect "1-run win%" to equal "2-run win%". I think we should expect them to be correlated, but that 1-run win% should be condensed. Something like:

1-run win% = ( x * 2-run win% ) + ( (1-x) * .500 )
if x = 0.5, then that would mean you'd expect your 1-run win% to be halfway between your 2-run win% and .500

I would guess if your 1-run win% equals your 2-run win% and both are above .500, then you were still somewhat lucky (and vice versa if you were a below .500 team).

I think in real-life baseball, something like that formula is right. 1-run win % is normally somewhere between (maybe half-way, I don't remember) .500 and overall win %.

But real-life baseball is definitely different to WIS. In real-life, exp win % is a much better way to predict win % going forward than 1-run record. (And looking at component stats like BaseRuns is better again.) But because of the weird role that MopUps have in WIS, I don't think that what we know about real-life should carry over to the game. As you said, the huge blowouts - I had a 47-11 win late in the season! - really throw off the exp win %.

The correlation coefficient between owners overall winning% to 1-run game winning% is .568.

The correlation coefficient between owners overall winning% to 2-run game winning% is .988

The correlation coefficient between owners 1-run game winning% to 2-run game winning% is .435

Nice, that sounds about right

Agree brianjw, the question is just how many of those blowouts are there and do they swamp all the other information you get from Run Diff.

Ideally, we would go to the game, or even the PBP level and de-emphasize the blowouts, but that's not possible.

We can go into the game level analysis, but not the PBP. About a day or so after the season ends, there is a link on the results page that says "Click here for the league spreadsheet." That gives you a zip file with a dozen or so CSVs. And one of them is the full list of games played, with the final score. You can use that to compute win-loss record, 1-run win record, expected win record. But you could also - and I might do this - see how many blowouts there are, and what happens if you reduce their import.

Doing this with the PBP would be better, but I don't think those records are kept.

I can't do this today - because not all the spreadsheets are up - or probably tomorrow because of work, but hopefully by the end of the week.

Annoyingly, the spreadsheets only stay up for a week or so, so you've got to be quick to grab them. But they are really helpful for understanding how the game works.

Amd for the record, it's not the 15-3 blowouts that skew the results. Those games happen all the time and are not necessarily caused by fatigued mop-up guys. It's the 25, 30 and 40 run games that screw up the results.

I totally agree. I was thinking of replacing every score over 15 with 15 and redoing the expected win %.

Posted by schwarze on 8/9/2020 5:54:00 PM (view original):
2+ Run Game Winning%
.
Look who's the leader in 2+ run games. Your fearless leader, ozomatli. The difference represents the winning% difference between 2+ run winning% and 1-run winning% A negative difference indicates bad luck. Sorted by 2+ Run winning%.
.

Owner	Actual Win%	1-Run W%	2+Run W%	Diff
--------------------------	-----------------	-----------------	-----------------	-----------
ozomatli	0.575	0.480	0.617	-0.137
06gsp	0.581	0.500	0.614	-0.114
skunk206	0.583	0.513	0.609	-0.096
justinlee_24	0.590	0.549	0.604	-0.055
mjkrunner	0.569	0.487	0.601	-0.114
schwarze	0.598	0.596	0.599	-0.003
brianjw	0.576	0.520	0.599	-0.079
jfranco77	0.580	0.545	0.594	-0.049
mensu1954	0.578	0.560	0.584	-0.024
d_rock97	0.552	0.479	0.582	-0.103
wink0094	0.557	0.504	0.577	-0.073
ejstockman	0.557	0.502	0.576	-0.074
marinerrott	0.545	0.460	0.575	-0.115
milest	0.566	0.549	0.573	-0.024
rbow923	0.546	0.488	0.571	-0.083
tigerrott	0.561	0.537	0.570	-0.033
hebdomad	0.538	0.450	0.570	-0.119
discodemo	0.562	0.545	0.569	-0.024
jbohrman	0.544	0.483	0.567	-0.084
ledfoot	0.542	0.484	0.565	-0.081
mildnhazy	0.549	0.515	0.562	-0.047
PennQuaker	0.555	0.538	0.561	-0.024
pkearns	0.560	0.576	0.554	0.023
magicdreamer	0.537	0.500	0.553	-0.053
besterateam	0.536	0.495	0.552	-0.058
buddhagamer	0.543	0.520	0.552	-0.032
beauchamp	0.534	0.495	0.551	-0.056
thehuseman	0.553	0.562	0.551	0.011
mpitt76	0.527	0.473	0.547	-0.073
bottomlee	0.520	0.452	0.544	-0.092
ballsweatsux	0.544	0.543	0.544	-0.001
Jtpsops	0.545	0.548	0.544	0.004
crystalao	0.530	0.500	0.543	-0.043
barracuda3	0.529	0.496	0.541	-0.045
wildthings1	0.520	0.463	0.541	-0.077
Humdogs	0.534	0.517	0.540	-0.023
cstrohmier	0.529	0.502	0.540	-0.039
doctorcc	0.551	0.584	0.539	0.045
dougpalm	0.541	0.551	0.537	0.014
teamhydro	0.529	0.513	0.536	-0.023
NebHusker	0.534	0.536	0.533	0.002
jetscanes	0.538	0.553	0.531	0.022
redcped	0.524	0.513	0.528	-0.014
skinndogg	0.524	0.525	0.523	0.002
BigScungil	0.501	0.444	0.522	-0.078
Fusion27s	0.513	0.495	0.521	-0.026
bheid408	0.514	0.506	0.518	-0.012
odalisgagne	0.521	0.539	0.514	0.025
thunder1008	0.498	0.457	0.513	-0.057
dorkster	0.517	0.529	0.513	0.016
newarkwilder	0.506	0.509	0.505	0.004
wetwillytwo	0.497	0.476	0.505	-0.029
sford	0.494	0.464	0.505	-0.041
duhbigcat	0.486	0.444	0.501	-0.057
daddyzander	0.495	0.482	0.500	-0.018
cubbies84	0.484	0.450	0.496	-0.046
just4me	0.502	0.526	0.493	0.033
footballmm11	0.498	0.522	0.490	0.032
mickr9	0.494	0.504	0.490	0.014
Jdh34	0.483	0.475	0.486	-0.011
mllama54	0.495	0.527	0.482	0.045
contrarian23	0.474	0.458	0.481	-0.022
pfan	0.480	0.481	0.480	0.001
npg32433	0.485	0.500	0.479	0.021
chargingryno	0.488	0.511	0.478	0.032
nordawg	0.463	0.428	0.476	-0.048
calhoop	0.486	0.512	0.475	0.038
Chisock	0.485	0.520	0.472	0.048
thebubbaq	0.480	0.510	0.470	0.040
elvisgump	0.467	0.475	0.464	0.011
markeking	0.480	0.528	0.461	0.067
midknight	0.468	0.494	0.459	0.036
DarthDurron	0.448	0.423	0.457	-0.034
trd3	0.471	0.519	0.454	0.066
Sonofodin52	0.460	0.500	0.447	0.053
stroh23	0.449	0.458	0.446	0.012
gorbyruss	0.464	0.518	0.445	0.073
tracyr	0.456	0.488	0.445	0.043
MitchellBade	0.464	0.516	0.443	0.073
loyalcanuck	0.451	0.479	0.441	0.038
big_dowg	0.466	0.546	0.437	0.109
happyhours	0.444	0.472	0.435	0.037
cholatse	0.439	0.461	0.431	0.030
toddcommish	0.444	0.493	0.425	0.067
garmansouth	0.437	0.485	0.419	0.066
sjh0825	0.418	0.426	0.415	0.011
gigrant	0.434	0.507	0.412	0.095
northof49	0.426	0.504	0.401	0.104
jschirn	0.403	0.444	0.390	0.055
bigsteve12	0.408	0.511	0.369	0.142
deathdealer	0.408	0.538	0.363	0.175
marvin_katko	0.368	0.432	0.345	0.087
tridentric	0.378	0.472	0.344	0.128
Zachsmikle1	0.356	0.436	0.332	0.104
spoonfed	0.353	0.443	0.324	0.119
steelerfan19	0.219	0.414	0.165	0.249

Knowing that I was on a 600-win pace outside of 1-run games is a tough pill to swallow.

If you ever miss the spreadsheet for a league admin can send it to you. Missed one of my leagues and they were able to provide it

Oh that's really good to know. Thanks!

I stopped collecting them a little while back because there was a period when the automatically produced ones were really incomplete. But they seem to be working again now, so hopefully I can run a little analysis on the leagues we all just played.

Posted by schwarze on 8/10/2020 3:01:00 PM (view original):
Amd for the record, it's not the 15-3 blowouts that skew the results. Those games happen all the time and are not necessarily caused by fatigued mop-up guys. It's the 25, 30 and 40 run games that screw up the results.

I decided to see how this would have affected me. One sample size, but it's definitely dependent on theme/cap:

80M: 0
90M: 9 (4 of them 20+ runs)
110M: 3 (none over 17)
120M: 0
140M: 5 (2 of them 20+ runs)
Var: 6 (1 of them 20+ runs)

I wonder if the 90M caused as much mayhem for others due to the defenses. But I find it interesting that the 80M and 120M, both of which played as lower caps really, didn't have any at all.

Posted by redcped on 8/11/2020 1:53:00 PM (view original):

Posted by schwarze on 8/10/2020 3:01:00 PM (view original):
Amd for the record, it's not the 15-3 blowouts that skew the results. Those games happen all the time and are not necessarily caused by fatigued mop-up guys. It's the 25, 30 and 40 run games that screw up the results.

I decided to see how this would have affected me. One sample size, but it's definitely dependent on theme/cap:

80M: 0
90M: 9 (4 of them 20+ runs)
110M: 3 (none over 17)
120M: 0
140M: 5 (2 of them 20+ runs)
Var: 6 (1 of them 20+ runs)

I wonder if the 90M caused as much mayhem for others due to the defenses. But I find it interesting that the 80M and 120M, both of which played as lower caps really, didn't have any at all.

Probably due to weaker offenses. Mopup guys are usually similar quality across all leagues, but offenses typically improve at higher caps, which = more damage.