Kilbins & Players & Rankings (Oh My!) (566 views)

[Deleted]

Aren’t you just as tired as I am about Ian taking winshare posts out of context in shout? The dude just can't seem to understand that player's don't play equivalent minutes per game, or games per sim. Don’t you wish there was a better way to rank teams without using fracs or SRS, which fall victim to season-long problems with injuries or trades changing the competition of teams? Want something simple that isn’t complicated but is still complex enough to calculate how good the team you are about to face is right now? Well have I got the ranking system for you!

I call them Kilbins, and the higher a team’s Kilbin value the better their roster is at the current moment. How is Kilbin calculated, you ask? Well, I won’t get into the details, but it involves looking at the current WS/48 of the players on your roster, and their projected minute distribution. This is then presented in units of “WS/48”, so you can view .110 as approximately league average. Note that the estimates are rough as we’ve only had a few sims, but they should get better later on.

First things first here is a fun little chart comparing a teams actual wins to the total winshare of the players on their team, you can see “total winshares of the players on a team” accounts for (R2 Coefficient) about 75% of a teams wins on average.

And now, the rankings from the last few years:

And now for this season:

As you can see the Lakers and the Chapparals are looking extremely good relative to the pack. The Nets, Bucks, and Bobcats round out the top 5. The heat, however, are really coming strong so far in the 6th spot. I’ve calculated the “expected wins” of a team with the currently assembled roster and compared them to actual wins, and show the difference of Wins - Total Team Winshares as “Luck”. This is the deviance from the 1:1 of each team you see in the plot above. Additionally I’ve calculated the “expected wins” of the team if no moves are made after todays date. This is calculated by looking at the Actual Wins of the team, and calculating how many more wins they will get based on how many games are left and their Kilbin rating.

And Finally, here are the winshare and winshare/48 statistics used for each player sorted by each team for these calculations! This isn’t necessary to share, but I figured having the WS/48 of players in the league handy is helpful since it isn’t shared anywhere on the site at the moment.

[eric]

great stuff bill

i think something is off in your ws/48 denominators, those values all look really low, maybe double check if you used minutes through sim 5 instead of 3 for that part?

i also think it's important to clarify one thing - you're not taking a team and adding up all the win shares of players only while they were on that team, but adding up the full season win shares of players currently on a team's roster. so for example, 76er players through sim 3 generated 21 win shares, but they're listed at 18 here for reasons like shawn kemp and his 3.5 being traded away. in short, the point of this analysis is not to see how close team win shares get to team wins, but to see if we can get a better idea of what a team's strength right now is, taking into account trades out and in

[Deleted]

Apr 13, 2021 16:52:43 GMT -5 eric said:

great stuff bill

i think something is off in your ws/48 denominators, those values all look really low, maybe double check if you used minutes through sim 5 instead of 3 for that part?

i also think it's important to clarify one thing - you're not taking a team and adding up all the win shares of players only while they were on that team, but adding up the full season win shares of players currently on a team's roster. so for example, 76er players through sim 3 generated 21 win shares, but they're listed at 18 here for reasons like shawn kemp and his 3.5 being traded away. in short, the point of this analysis is not to see how close team win shares get to team wins, but to see if we can get a better idea of what a team's strength right now is, taking into account trades out and in

Yes it is the sim 5 vs sim 3 thing, I had started this article the other day but had to re-pull the data and unfortunately the stats from sim 3 are lost to time.

to the second point, yes. Thats why the trend value is only like .75 instead of way closer to 100. So the article is a little bit of a two parter. The ranking system is a separate from the trend graph, but the trend graphs are fun. Its amazing how well it does even when a lot of trades/roster shuffling happens late in the season.

[Soup]

Great job Bk. You put a lot of work into this.

100k

[Handsome Pete]

[Deleted]

I'll update the Kilbin ratings tomorrow for Sim 6 winshares/minutes to fix the scaling, and also update all the graphs again at the end of the season.

[20s]

90k

[jhb]

Do you think it’s luck or skill when someone consistently exceeds their team WS/48

[jhb]

100k

[Sapular]

100K. Great thinking here

[Deleted]

Apr 13, 2021 18:55:57 GMT -5 jhb said:

Do you think it’s luck or skill when someone consistently exceeds their team WS/48

its coty material

[pointyegg]

100k

[delap]

69k

[skrouse]

75K - looks like we've acquired at least 2 pieces for our long term future. Not bad

[Deleted]

Kilbin ratings updated for sim 6

[bankz]

Chaps on pace for 83 wins

[Deleted]

ratings updated... i also have changed the name of "expected wins". Its more accurately Wins + Expected Winshares. As we can see winshares typically exceed wins... maybe in a few seasons I will feel comfortable doing a simple Expected Wins = a*Expected Winshares where a is an empirically derived constant but I don't want to do that yet

[Deleted]

First off, everyone's winshares and winshares per 48 by team. Let me know if you see any errors!

[Deleted]

And here is the final regression of winshares vs wins. A lot of correlation! I bet my earlier charts were plagued by the all-nba team name bug.

[jhb]

Add an extra digit to the WS/48 column you heathen

[Deleted]

And finally... I re-did Kilbins for the 3003 and 3004 season. Things are a lot more bunched up this year, which I think everyone would agree with anecdotally. Cougars, Warriors, Nets, Sixers, Chaps all have great shots according to Kilbins. The 2nd tier are the Bucks, Pacers, Bulls, Timberwolves.

[20s]

@bk Butch Carter shows up under my team and Druce’s. He’s not on my team.

[eric]

you can also do custom format in excel to get rid of the leading 0.s

[Deleted]

sorry 20s thats just a c/p error different sheets entirely from the kilbins so no error there.

thanks eric

[andrewluck]

Apr 13, 2021 17:04:30 GMT -5 Soup said:

Great job Bk. You put a lot of work into this.

100k