BuzzerBeater Forums

Go through any number of games you like, look at the number of points scored and the number of PP100 predicted for players who attempted zero free throws, and then see how many times that the PP100 is exactly equal to points divided by shots multiplied by 100. It will be exceedingly rare for that to be the case, because that's not at all what it is measuring. If that were the case, they'd have it down in the stat line with steals, points, blocked shots, etc. Instead, it's up with the team ratings, which are also abstract calculations by the engine that are not direct results of what happened in the game, but instead a tool to evaluate how your team performed.

Υοu've lost me there.
I've done the calculations with FTs included as attempts( 2fts= 1AFG), i was getting way off the mark(final score or points scored per position).
After ΒΒ-Μarin's clarification i excluded free throws, was still getting as you say too, big inconsistancies.
But i can't pinpoint where exactly you disagree with me.

The team ratings can be calculated. It is stated they have nothing to do with the opponent, only our players skillsets, game shapes, enthousiasm, location and lately GPD prediction.

Go through any number of games you like, look at the number of points scored and the number of PP100 predicted for players who attempted zero free throws, and then see how many times that the PP100 is exactly equal to points divided by shots multiplied by 100. It will be exceedingly rare for that to be the case, because that's not at all what it is measuring. If that were the case, they'd have it down in the stat line with steals, points, blocked shots, etc. Instead, it's up with the team ratings, which are also abstract calculations by the engine that are not direct results of what happened in the game, but instead a tool to evaluate how your team performed.

Υοu've lost me there.
I've done the calculations with FTs included as attempts( 2fts= 1AFG), i was getting way off the mark(final score or points scored per position).
After ΒΒ-Μarin's clarification i excluded free throws, was still getting as you say too, big inconsistancies.
But i can't pinpoint where exactly you disagree with me.

The team ratings can be calculated. It is stated they have nothing to do with the opponent, only our players skillsets, game shapes, enthousiasm, location and lately GPD prediction.

Mostly the bold stuff:

As to hrudey's explanation, it's hard to accept without official verification.
Mainly because pp100s relate to the game's attempted shots. They should.
After taking into account all the conditions these shots were taken under, then the GE should calculate the equivalent points in 100 similar shots.
Otherwise, what does it calculate? How many points we would get a hundred times a player touched the ball or had the ball and was facing the basket etc? It doesn't make sense.
The rating is points per 100 shots(similar to those taken in the game). If there are no shots, there should be no pp100.

Οk, im back on track

So, in your opinion if the pp100 doesn't relate to the game's attempted shots, where does it? Or you believe that it's not a necessary requirement, they do relate (only)to something "existing/identifiable"?

Imagine coin flip.
1st day coin was flipped 100 times. Result: 60 heads, 40 tails.
2nd day coin was flipped 2 times. Result: 0 heads, 2 tails.
3rd day coin wasn't flipped. Result: 0 heads, 0 tails.
What can we say about boxscore?
Boxscore in 1st day: 60-40.
Boxscore in 2nd day: 0-2.
Boxscore in 3rd day: 0-0.
So what can we say about probability (heads and tails per 100 flips)?
Probability in 1st day: 60% - 40%
Probability in 2nd day: 0% - 100%
Probability in 3rd day: non exists.
Wrong. Even children know that coin flip probability in all three days always is 50% - 50%.
Probability exists even without attempts. And result of limited number of attempts very often is different than probability.
So it is what points per 100 shots means in BB.

Of course PP100 is broken (or at least misleading) because it doesn't include team effort, but it is issue to other discussion. So when you are CTed then PP100 means "theoretical PP100 if opponent wouldn't play CT" or something like that.

Last edited by B.B.King at 1/22/2014 4:29:17 PM

Are yoy telling me that GE is "flipping a coin", in order to produce the pp100s and doesn't calculate each match's constants and variables in order to come up with a result?
I seriously doubt it, there is no indication anywhere, neither in the game manual nor any BB statements, for such thing.
Im ok if told officialy that this is the case.

Common belief is that pp100s include all aspects of the match ups. Opponents skills, stamina, enthusiasm, attitude, location, the whole lot. Except FTs.
Having that in mind, all the differences from my calculations to the final points scored, i attributed them to "randomness".
But even randomness would have a very large margin, from game to game.

My suspicions are two.

a) Randomness exists as calculated, very unpredictable with regards to its peak point. This theory intergrates the probabilities you mentioned.

b) pp100s are very accurate, BUT one or more of the variables that we believe to be intergrated, actually isn't.
Since we get almost always, in the same match, some players from the same team overscoring and others underscoring(pp100 wise), such variable wouldn't be enthusiasm, attitude or location for instance, that affect a team's players equally.

Are yoy telling me that GE is "flipping a coin", in order to produce the pp100s and doesn't calculate each match's constants and variables in order to come up with a result?

Nope. PP100 is calculated by formulas. Result of match (boxscore) is like flipping a coin.
In simple flipping a coin formula is easy, always 50% - 50%. Result can be 60-40 or 0-2. This difference between probability and result of limited number of attempts. When number of flips goes to infinity then result goes to probability.
PP100 is like probability, boxscore is like limited number of attempts.
So we have some probability, and then random final result is generated. So You can't calculate probability based on random final result.

Common belief is that pp100s include all aspects of the match ups. Opponents skills, stamina, enthusiasm, attitude, location, the whole lot. Except FTs.

For sure PP100 doesn't include team effort.

Οk, thanks for the clarification(and the patience), you made it absolutely clear to me.

So, you're advocating the theory of small numbers randomness. Makes total sense and is the most likely scenario.
If this is the case, what annoys me, is the (very precise)numeric valuation of the pp100, as a match up rating. It throws you off, straight away. Very misleading imo.

They were too confusing, to put it simple:

PP100 has nothing to do with what actually happened in the game.

Yer its not a straight forward concept to begin with - BUT

once you understand it, it helps you make a lot of decisions and analyse the efficiency of your team.

Lets take your example where the C had a PP100 of 51. I dont know how the rest of the team faired, but lets say, that the rest of your team had better PP100 ratings.

Now lets say hypatheticall, that you believe that your C is actually really good offensively, and you play Low post/Look inside consistently.
Now maybe, in other games, ur C's PP100 is up around 100, 120, which is pretty good. yet this one game, it was down to 51. You know what that means? The opposition C has AMAZING defence against ur C

So next time you play that team, you can make better assessments about the type of offence. Or play tactically smarter - stick your gun C at PF where perhaps his opposition PF is not as good at defending.

Stuff like this. - It's a really good tool for analysing games.

It is a good analysing tool indeed, with many different aspects to weigh in order to reach a conclusion.

My objection is that, when the game uses a known statistical category, with numerical(to the decimals) valuation, you expect it to be accurate and reflect facts.
If thats not the case, there should be a different kind of valuation(i.e good,bad,excellent etc), or keep the numerical one but clearly state the fact that it's not accurate and also state the factors that may contribute to that.
We will still have to assess these factors and also the pp100 results, which is not straight forward anyways and exciting imo.