Search found 23 matches

by vzografos
Sat Jun 15, 2019 5:50 pm
Forum: APBRmetrics
Topic: Per Game vs Per Minute?
Replies: 2
Views: 458

Re: Per Game vs Per Minute?

It's popular these days to go per-possession or per 100 (or 200 if you count defensive). But possessions are estimated and/or averaged, or for whatever reason different for team and opponent; and some things -- turnovers? -- probably are more minute-related than possession related. "Per 36 minutes"...
by vzografos
Sat Jun 15, 2019 11:12 am
Forum: APBRmetrics
Topic: 2019-20 team win projection contest
Replies: 12
Views: 1387

Re: 2019-20 team win projection contest

Not a stat head, so can't help too much on the details, but for the 10 player bit I would say that you may be able to get away with a lower number than that (if that at all simplifies it, it's possible there's no real difference between running it with 1 or 100), in my experience you can usually ex...
by vzografos
Fri Jun 14, 2019 10:12 pm
Forum: APBRmetrics
Topic: Per Game vs Per Minute?
Replies: 2
Views: 458

Per Game vs Per Minute?

So I am looking at Player stats (from stats.nba.com) and I was wondering what would be the best stats to train my model on, considering that I want to avoid statistic lag (this is the time it takes until average statistics of a player stabilitse in the season), which causes a lower prediction accura...
by vzografos
Fri Jun 14, 2019 9:35 pm
Forum: APBRmetrics
Topic: 2019-20 team win projection contest
Replies: 12
Views: 1387

Re: 2019-20 team win projection contest

I haven't tried to read or follow the links to method. I'll just ask are the "chances to win" just averaged or do the weights vary by player? I'd think they'd have to be variably weighted. So given that each player in the team has a different win chance (lets call this metric Wn, which can be calcu...
by vzografos
Fri Jun 14, 2019 6:50 pm
Forum: APBRmetrics
Topic: 2019-20 team win projection contest
Replies: 12
Views: 1387

Re: 2019-20 team win projection contest

Incidentally, if someone wants to go for the single-dimension predictor (e.g. the PIE score for each player that is on the stats.nba.com) you can try the following: -Derive a single metric that ranks each player's chance of win. This metric can be estimated as a function of different player paramete...
by vzografos
Fri Jun 14, 2019 6:28 pm
Forum: APBRmetrics
Topic: 2019-20 team win projection contest
Replies: 12
Views: 1387

Re: 2019-20 team win projection contest

Ok I see.
Well for my predictions I use the minutes as an additional parameter/feature and not collapse everything into a minute-weighted dimension.

I also found that a combination of player-based and team-based features works the best
by vzografos
Fri Jun 14, 2019 8:17 am
Forum: APBRmetrics
Topic: 2019-20 team win projection contest
Replies: 12
Views: 1387

Re: 2019-20 team win projection contest

Anybody want to work collaboratively on a freeshare minutes model? Say in August? What else would improve predictions? Anybody want to swap model building techniques, in public or private? Anyone want to try a team entry as a novelty? Depends how you prioritize winning over learning and fun. Hi for...
by vzografos
Fri Jun 14, 2019 8:16 am
Forum: APBRmetrics
Topic: Reconstructing Box Plus/Minus
Replies: 99
Views: 41085

Re: Reconstructing Box Plus/Minus

Gradually working on this project! Yes, the idea is to maintain the existing general concept of BPM (i.e. historic applicability, general structure) and significantly improve the handling of outliers. Thus far I have focused on the linear version of BPM, currently called GmBPM. It should be very st...
by vzografos
Thu Jun 13, 2019 3:21 pm
Forum: APBRmetrics
Topic: Reconstructing Box Plus/Minus
Replies: 99
Views: 41085

Re: Reconstructing Box Plus/Minus

Box score stats only (i.e. anything that can be calculated from the stats we have from the 80s.) No PbP stats, not even things like "assisted by" ratios. Nothing super complex that can't be done by someone with Excel and a good knowledge of math. Focus on Explanation, not Prediction. What happens s...
by vzografos
Thu Jun 13, 2019 1:13 pm
Forum: APBRmetrics
Topic: 2018-19 team win projection contest
Replies: 154
Views: 134453

Re: 2018-19 team win projection contest

Mike G wrote:
Thu Jun 13, 2019 12:12 pm
Your predictions are set in stone before any games are played, and they are not updated.
I see I get it.

Regarding the error metrics?

Also, are previous winners going to give us some hints/indication of their approach? :mrgreen:
by vzografos
Thu Jun 13, 2019 1:11 pm
Forum: APBRmetrics
Topic: How did you do for 2018/2019 regular season?
Replies: 8
Views: 783

Re: How did you do for 2018/2019 regular season?

That dip at 30-50% of the season is likely due to these providers' insistence on regressing to their prior expectations. As the season wore on, it became unavoidably clear the Clippers and Bucks and Blazers were actually doing this well, and others were going to fall below pre-season predictions. D...
by vzografos
Thu Jun 13, 2019 6:29 am
Forum: APBRmetrics
Topic: 2018-19 team win projection contest
Replies: 154
Views: 134453

Re: 2018-19 team win projection contest

Well joined the party late (as usual). I'll be watching this thread for now and maybe participate in the 2019/20 season. Albeit just for the regular season. Was wondering though, your Abs. Err, and RMSE metric are in relation to what exactly? Are you looking at wins/losses (accuracy) only or also pr...
by vzografos
Thu Jun 13, 2019 6:02 am
Forum: APBRmetrics
Topic: On Datasets...
Replies: 0
Views: 412

On Datasets...

Ok so I know this topic has been discussed before, some years ago, but I was wondering what kind of datasets are you guys using for your predictions nowadays? Is it datasets you built yourself, maybe you downloaded from somewhere or you bought? Just curious as to what everyone is using. I have scrap...
by vzografos
Thu Jun 13, 2019 5:33 am
Forum: APBRmetrics
Topic: NCAA->NBA ML using raw text of historical scouting reports: done!
Replies: 20
Views: 2709

Re: NCAA->NBA ML using raw text of historical scouting reports: done!

The forecast variable is an average of three win production measures (Win Shares, Wins Produced, and Estimated Wins Added) over each player's three years after the draft, or zero for years they did not play. Interesting model/approach. Ok I do not really follow (or claim to understand) NBA drafts, ...
by vzografos
Thu Jun 13, 2019 4:47 am
Forum: APBRmetrics
Topic: How did you do for 2018/2019 regular season?
Replies: 8
Views: 783

Re: How did you do for 2018/2019 regular season?

https://i.ibb.co/RNZVD7V/untitled.png So here is a graph showing the prediction accuracy for each of the different providers. The accuracy is accumulated after each game, with a total of 1230 games in the regular season. Observe the noisy nature between 0-200 games, which is quite expected for any ...