Predicting Raptors Wins With Wins Produced

[malefax]

"Wins Produced" is a stat that tries to measure the contributions each player makes towards the number of wins a team makes. Although I think there are some problems with using it to allocate credit for winning to individual players (the main thing is that it gives rebounders a lot of credit, and can't measure individual defense) one thing it does extremely well is predict how many wins a team as a whole will get in a year. So, for example, last year if you add up the "Wins produced" by all the raptors players, it comes out to 36.5 or so, which is pretty darn close to 40. And this is typical.

So one way of predicting how many wins the Raptors will get is to figure out how many wins their players will produce, and add those up.

So lets do that. For this we have to:
1. subtract the players who left.
2. Reallocate their minutes to other players
3. Figure out if there are any likely changes to the productivity of players
4. Do a bit of math

So, 1:
Bosh produced 14 wins and played 2526 minutes.
Turkoglu produced 4.2 wins and played 2272 minutes
Wright produced 0 wins and played 1392 minutes
Bellinelli produced 0.6 wins and played 1121 minutes.

So 36.35 wins - 18.8 wins gives us: 17.55 wins

Those are the major subtractions. This leaves
2526 extra minutes for bigs and
4785 extra minutes for wings

2. obviously we can't predict injuries, but here goes;
Johnson +726 big minutes (bringing him up to 26.5 minutes a night)
Davis 1000 big minutes (12 minutes/night)
Kleiza 2200 minutes (200 as a big)
Barbosa 2000 minutes (24 minutes/night)
DeRozan +400 minutes (2064 total, about the same as Barbosa)
Weems +385 minutes (1718 total, 20 minutes/night)
Evans +500 big minutes (811 total, about 10 minutes/night)

Obviously there are a lot of guesses there, but bear with me for a moment.

3. What productivity changes should we assume?
I will assume that Bargnani, Johnson, Calderon, Jack, Evans, and Weems will be equally productive this year as last year.

Then I will make two sets of assumptions: one pessimistic, one optimistic.
Pessimistic: Kleiza's WP/48 (that is, how many wins he produces per 48 minutes) is equal to what it was in Denver
Barbosa's productivity is about his career average (I assume .08)
Davis's productivity is .05 (well below average)
DeRozan's productivity is the same as Weems was last year (it seems almost certain he will improve, since rookies almost always do)

Optimistic: Kleiza's rebounding and offensive efficiency will be closer to what we saw in turkey and the Euroleague
Barbosa's productivity will be similar to what we see in his good years in Phoenix (.12)
Davis's productivity is average for his position (not unlikely if he rebounds well)
DeRozan's productivity is average for his position.

4.
Pessimistic:
Johnson 8.85 extra games worth of minutes x .168 wins/game = +1.5 wins
Kleiza 26.8 games x .096 wins/game = 2.5 wins
Barbosa 24.4 games x .08 wins/game = 2 wins
Davis 12.2 games x .05 wins/game = 0.5 wins
DeRozan 25.2 games x .075 wins/game = 2 wins (this has to be added because last year he had no wins due to lower productivity, we are assuming)
Weems 4.7 games x .075 wins game = 0.5 wins (rounded up from .3)
Evans 6 games x .116 wins/game = 0.5 (rounded down from .7)

That adds up to 9.5 wins, which gives us 27 wins on the season.

Optimistic
Johnson: same +1.5 wins
Weems and Evans same, +1 win total
Kleiza: .15 wins/game = 4 wins
Barbosa .12 wins/game = 3 wins
DeRozan .1 win/game = 2.5 wins
Davis .1 win/game = 1 win

That adds up to 12.5 wins, which gives us 30 wins on the season. Not a big difference, eh?

So what's the point of all this math?

I think it tells us that we are about a 30 win team if people play about as well as they have in the past, even if DeRozan picks up his play a fair amount. Basically, for us to make the playoffs, something unexpected will have to happen.

In order for us to do better, someone playing a lot of minutes will have to pick up their game. I think this is distinctly possible, but not something we can count on. The major possibility is Bargnani. I don't want to say it's a straight causal relation, but if he starts producing at a less awful rate, that could impact big in the win column.

Tweet

[Buddahfan]

You have to adjust the Wins produced for the allocation of usage not just minutes.

Bosh had the highest usage per 36 minutes of all Raptors players.

So you can not allocate just his minutes.

1. You have to allocate his usage along with his minutes

2. Also there is no way to extrapolate Kleiza Wins Produced while at Denver was he was strictly a backup role player to how he will be used by the Raptors

3. Then you have to take into consideration that players like DeRozan will up their WS/48 over his rookie season.

4. You also need to take into that Bosh was endothermic and Johnson is exothermic when it comes to the other Raptors players on the court. If you add up all the Win Shares of the Raptors players or other teams it comes to less than the total team Wins. If you look at the Cavs that number is extreme. That is because players like Lebron contribute more to their team winning than just what you see in the numbers. The difference on the Raptors was not near that large because Bosh is endothermic. The difference in those two numbers with respect to the Raptors will be larger in 10-11 than it was in 09-10 for a number of reasons including the fact that in my opinion Bosh does not make the other four Raptors on the court better despite his being double teamed. Johnson on the other hand will make the other four Raptors on the court better in 10-11 just he did in Detroit and did in Toronto last season.

So there should be a net gain in actual team wins over Wins Produced in 10-11 compared to 09-10.

5. There could be some other issues with your analysis so my list is not meant to be inclusive.

Bottom line is that it seems to me that there is room for improvement in your assumptions.

Nice try but back to the drawing board.

[TRX]

It's true malefax forgot to take into account USG% as well as minutes, but re-allocating Bosh's usage to role players will, in fact, lower most of those players' efficiency and WS/48. How the two effects (increased minutes and increased USG%) counter each other has obviously yet to be seen. I agree with your #2 point, however. Kleiza will likely see a tremendous change in role from more-or-less just a spot up shooter, playing average defence to being more involved in the game on both ends of the court. Point #4 is purely speculation - not just by you, but by every prediction... which is probably why it's not part of any statistical analysis until after the fact.

If I recall, Buddah, you tried to disguise your own personal Raptor optimism under the cloak of objectivity by the same argument (chemistry should add another whatever wins to the modest total the actual numbers point to) in a previous thread you started. Not that you're incorrect... but it's, by definition, not in the name of statistical analysis.

[malefax]

Buddahfan:
1. Not really. The effects of rebalanced usage are likely to be insignificant. Usage doesn't have a huge impact on WP48 to begin with (unlike, say PER), and both Kleiza and Barbosa are relatively high usage players that will soak up a bunch of those extra possessions.

2. On the contrary, there are lots of ways to extrapolate Kleiza's WP48. I used 2 different ones. One of the things that the discussion shows is which one you use doesn't really matter much. The only way Kleiza could really throw things off a lot is if he gets a lot of minutes AND is at the top end of the plausible WP48. In that case, he could be worth 7-8 wins instead of 3-4

3. If you read the post, I DID take into account that players like DeRozan will up their WP48. His WP48 last year was basically 0.0

4. That's missing the point. The point is to try and come up with some sort of number based on unbiased quantitative evidence. If you think the Raptors will be more exothermic this year, fine, then you take the estimate of 27-30 wins and instead of adding 4, add 7 or 8. How much you add will depend on your subjective opinion on the topic. But since we can't use WP48 to predict these interaction effects, we obviously can't factor it directly into an analysis that uses wins produced.

5. Of course there are other issues, but so what? The question is, are they significant enough to unbalance the estimate? And I can answer that question a lot better than you just tried to. The major things that will mess up the estimate are if the minutes are significantly different for some people, or if productivity of anyone changes a whole bunch. The minutes thing is tough but, for example, if Johnson plays 36 minutes a night that's worth an extra win or two, etc. And the other issue is the error between the wins as predicted with win score and actual season wins. That is less of a worry for me, just because the error is statistically not that huge. So although Wins Produced has all kinds of imperfections as a model of reality, we can still be pretty confident that estimates using it won't be too far off.

[Edgar]

I like it - It seems like a rational and unbiased use of advanced statistics.... I would almost guarantee that we end up 5 games over or under those estimates...

[Gman]

The concern I have with this is you have an a** sucking Hedo adding more wins per than any of the new Raptors coming. I understand that good Hedo might have done that but bad Hedo shouldn't even be close to that. Either he didn't help us winning as much or others will help it more.

[malefax]

Hedo racked up assists and rebounds at a adequate rate, and he played a lot of minutes, hence the 4 wins. If Barbosa or Kleiza plays as many minutes as Hedo did, and they are having a decent year, it shouldn't be too hard for them to get 5 or 6 wins (i.e. more than him). But I assumed they would each be playing a fair amount fewer minutes than him, because he still have to make room for more minutes for DeMar and Weems as well (I assume).

[SirChillyMost]

Remember Jay is the coach how do you stat that up in relation to projected Raptor wins????

[OzRapFan]

Remember Jay is the coach how do you stat that up in relation to projected Raptor wins????

Divided it by 2.

[PapiJulio]

How can people disagree on speculations? Just for the sake of argument I guess.

'Malefax' I found your analysis a pretty cool insight on a dismal up coming season. Too many Raps fans are in such a negative mood because of the current roster. I say what's the point in making the playoffs? It's all about developing players and setting individual goals this year.

You must be quite a fan to spend that much time crunching numbers for the Raptors. I hope the players will give that kind of effort this year! Who knows, maybe will beat everybody's expectations and win a whopping 31 games! BUT I hope we come dead last. Anything more would be a disappointment. We could pretend it's like golf this year..... lowest score wins... GO Raptors GO!

[minks77]

Wow. A whole bunch of number crunching and time spent to come up with a prediction that means less than nothing. You guys have far too much time on your hands. People love to diss the box score but it at least tells you who did what in which category night in and night out. Projections based on Win share, per 36 etc are entirely meaningless. This isn't to say that math can't explain what has happened in the past but predictions are best left to crystal balls.

[malefax]

Entirely meaningless? Pfah! Too much time on my hands? Pfah! If you're so busy, go do something productive with your time, Mr. Scoldsy McHavenofun

also, Win Share is based on box score stats so you officially make no sense

[minks77]

I can base the prediction of the weather on what was recorded in the past but that doesn't mean that just because last year was nearly snowing on Nuit Blanche that I'll just grab my parka tonight. So there, it makes perfect sense. Using stats to predict future outcomes is very tricky and usually a fools errand which is why they play the games on the court and not on paper. Besides, there is this thing called growth, ie as you age you grow, mentally, physically, spiritually etc. What happened yesterday may help give you an idea of what may happen tomorrow but it isn't going to give you anything more than that.

And I'm trying to wake up without any coffee so right now I don't have anything better to do than be cranky and poo on your "advanced anal-ytics." It's a pet peeve of mine, kinda like the offseason.

[malefax]

As per usual, the complaints (here from minks77) about doing some statistical analysis comes down to: "Well there are lots of bad ways to use stats, so using stats is bad". Also, because making predictions is difficult and subject to error, therefore it is pointless. Only one of those arguments is a straightforward logical fallacy, but neither is any good.

For myself, my pet peeve is people who express strong opinions without having any evidence for them. Like, for example, declaring that the Raptors will DEFINITELY win x games, full stop, no evidence given. Or saying that Win Score is 'entirely meaningless' based on.... uh... hmm. I'll be the first to admit (or, rather, insist) that Win Score has flaws, but 'flawed' is a lot different than telling us 'less than nothing' (how does that work, btw?)

Using some analytics at least gives us a starting point for discussion that is pulled out of... thin air.

[YKR23]

LOL this was...interesting
but you have waaaaaaay to much time on your hands...