View Full Version : Accu-Stats break statistics

Fran Crimi
02-25-2011, 10:52 AM
Awhile back in response to Sofasnapper's post (copied below), I said I would call Pat Fleming to ask about the break stats. I spoke with him today and here's what he told me:

The following break stats taken from recorded Accu-Stats matches apply to players,both pro and semi-pro who break hard and don't play position with the cb:

Players who broke --- won 55% of those games.
Players who didn't break --- won 60% of those games.

Pat also said that the statistic did not take into consideration the difference in skills of the players playing against each other. This was an overall statistic taken from as many matches as possible, which included matches where the skill level between the two players was substantially different, as well as matches were the level of play between the two players was similar.

As stated above, players who played position off of the break like Archer, and those who utilized the soft break like Deuel were not included in the statistic.

<span style="color: #3333FF">Quote Sofasnapper: Here's a claim from a team member that he recently got from Harvey Mason, from whom he takes pool lessons:

Supposedly, using the entire Accustat database of 9-ball games, and over a lot of time, pros overall are 56% LOSERS on their own break. (Or have a 44% chance of winning a game they break, which is the equivalent).

The further claim is that there are only two exceptions for anyone meeting a standard of enough games in the database-- Archer and Strickland, and that THEY come in at a bare 52% winning percentage in 9-ball on their breaks.

I find this hard to believe. However, we know bad things happen on breaks. If you break dry with an open rack, likely a pro opponent cleans the table. If you scratch, ditto. If you make a ball and have no shot, there is the whole push out conundrum which might go either way. And etc.

I remember a couple of years back, Shane's first US Open, when Keith McCready had his 4th place or 5th place run, and they met on the tv table. Shane was the betting favorite, and Keith never made any ball on his break, and still won that match fairly handily (?).

Now it's POSSIBLE that the break STILL is helpful rather than hurtful, if the percentage of games won when not breaking is lower still than this number (say, 44% average wins when not breaking). But that would seem to contradict that if they win, on average, 46% on their break, their OPPONENTS who aren't breaking should be picking up that 54% winning figure.

What is your reaction? And can Accustats data be looked at by outsiders in some easy format? I try to believe impossible things everyday, a couple before lunch, usually, but on this one I certainly reserve my opinion without further info.


02-25-2011, 11:51 AM
Interesting information...but please verify your percentages...55% win and 60% don't win??? Something doesn't add up....or am I misunderstanding something?


02-25-2011, 01:45 PM
Thanks for following up with Pat, Fran. Interesting, but in a way that makes my head hurt! I'm in the same boat as Steve, wondering how this is even possibly true?

02-25-2011, 02:03 PM
I wonder what the numbers look like by the set. Maybe there is a large swing in the numbers that we dont see. I suspect there are days where when a players break is "on" he might win 70% of those games where they broke. But more often than not, would only break with mixed results, say in the 40% win range. Maybe a scratch thrown into any set could drag the average down below 50%.

02-25-2011, 02:47 PM
Hmmmmmmm -- thinx.
If players with a very good break won 60% of their breaks.
And if players with a not-so-good break won 40% or 45% or 50% or 55%.
And if the breaks were alternate, ie shared.
Then, it would be possibe for the stats to show a 50%/50% win/looze. Beleev it or not.
Thats the way arithmetik works.

Fran Crimi
02-25-2011, 07:23 PM
Steve, I asked Pat the same question. Apparently, those percentages do not necessarily have to add up to 100%.

Fran Crimi
02-25-2011, 07:29 PM
I'm sure you're right. Individual set statistics might be really different than the average. Think of all the sets that went 11-1 or 11-2.

Rich R.
02-26-2011, 08:49 AM
I have not been schooled in the science of statistics.
Can anyone say whether these numbers are really different enough to be considered statistically significant as opposed to a 50-50 coin toss?

02-26-2011, 03:13 PM
The statistical significance of the difference is partially based on the number of observations.

The usefulness of the difference is based on the number of observations and a few other concepts.

In general, it can probably be assumed that the stats are based on a sufficiently large sample size to be robust at this level of difference. In other words the stats are useful and probably present a good idea of what is going on.

I suspect this is true because of the large number of observations and the relatively (statistically) large differences.

02-26-2011, 03:20 PM
Hiya Joe.
U are good with stats. I am wondering. Would it be possible for players with a good (power) break-shot to win much more than 50% of their break games, and for players with a poor break-shot to win much less than 50%, and yet hav stats that showed (truly) a 50/50 win/loss for breakers (overall).
Would that be possible???????????

02-26-2011, 04:56 PM
Sure if power breakers are not breaking 50% of the time.

02-26-2011, 07:31 PM
Joe -- I am now cooking up some smart (i think) thinking about stats re breaking etc. But i am having trouble getting started, koz (the aforementioned) 55% and 60% dont add to 100%.

Each game played haz 1 winner & 1 loozer. Each game haz 1 breaker & 1 non-breaker.
So, if breakers win 55 of 100 games, then non-breakers must hav won 45.
What hav i missed????

02-27-2011, 02:08 AM
There are different samples for the stats given

02-27-2011, 11:55 AM
<div class="ubbcode-block"><div class="ubbcode-header">Originally Posted By: JoeW</div><div class="ubbcode-body">There are different samples for the stats given </div></div>

And that makes me question the results. In order to get an accurate average, wouldn't you need to use the same data for the results on either side of the equation?


02-27-2011, 04:20 PM
U dont even need the other side of the equation.
If the breaker won 55 out of 100 -- then he/she didn win 45.

02-27-2011, 04:29 PM
Anyhow, what the stats didn show (& probly karnt ever show) woz that a good power-breaker might win say 60% of hiz/her breaks.
Here we are talking about a group of players of equal ability, xcept that say 1 in 10 are good power-breakers.
I suppoze that the players with an average sort of power-break might win say 52% of their breaks.
And players with a poor break might win say 47% of their breaks.
But i havnt krunched the numbers yet. It wont be too diffikult to write an excel program that would work this sort of thing out.

Anyhow -- do u think a good breaker kan win say 60% of their breaks?????
Az i say, i think that ordinary stats wont (karnt) show this -- ie ordinary stats are sort of useless.

02-27-2011, 06:46 PM
Tough to really draw any conclusions from this. An average breaker with a great safety game could also end up with a great win percentage. Also consider that most players would not volunteer to give up the break(reagrdless of the stats).

02-27-2011, 08:29 PM
It might be good to hav a listing of players and break-games won. A ranked listing (ranked by %).
To inklood a column showing % of non-break games won by that player. And a column showing that ranking.

Plus a ranked listing (by %) of players and non-break games won.
To inklood a column showing % of break games won by that player. And a column showing that ranking.

Halfway down thems ranked lists u will see a player whoze performance iz the median performance for that list.
The median might be more meaningfull than the average i think.
The %'s relating to the median player will tell u what the average player might expekt to do -- whereaz averages only tell u what the average performance iz, not az meaningfull i think.

02-28-2011, 02:06 PM
This could be assessed with a Chi Square test. Type of Breaker (Power or Not power) X Break Shot (Yes or No) using won or loss (1 / 0) as the dependent variable. The games won and the percentages are calculated for each cell in the four way table. These results can be assessed for significance. No distributional assumptions need to be made. Lets say we have 20 games that can be classified. The cell contents are the number of times the person in the cell won the game.

Type of Breaker
... P NP
.B 7 8
NB 3 2

You can see that Power Breakers win 35% of the total games played when they break. In this hypothetical data there is a main effect for Break Shot and probaly no difference between the type of breaker (Power or No Power).

To determine if the differences are "real" we would need many more observations probably 200 or so.

If we look at only Power Breakers they won 7/10 games (70%) while Non Power Breakers won 8/10 games (80%) of their observed games. But this type of analysis is wrong. We have to use the total sample size (20) games to get a meaningful analysis.

The results presented by Fran suggest to me that the breaker is more likely to win regardless of the type of break used. If there are thousands of observations from AcuStats the conclusion that "breakers win more often" is probably reliable

02-28-2011, 04:31 PM
I am thinking that nonetheless no stats kan show whether a very good breaker haz a giant advantage.
Very good meaning praps very powerfull or praps having a very skilfull or akurat or cunning (power) break.
When i say karnt, i mean it would be extraordinaryly diffikult, probly needing special equipment even.

02-28-2011, 05:06 PM
<div class="ubbcode-block"><div class="ubbcode-header">Originally Posted By: Fran Crimi</div><div class="ubbcode-body">Steve, I asked Pat the same question. Apparently, those percentages do not necessarily have to add up to 100%. </div></div>

Was he able to explain it to your satisfaction?


02-28-2011, 05:53 PM
<div class="ubbcode-block"><div class="ubbcode-header">Originally Posted By: cushioncrawler</div><div class="ubbcode-body">I am thinking that nonetheless no stats kan show whether a very good breaker haz a giant advantage.
Very good meaning praps very powerfull or praps having a very skilfull or akurat or cunning (power) break.
When i say karnt, i mean it would be extraordinaryly diffikult, probly needing special equipment even.

All that is needed here are good definitions of the terms and reliable methods for classifying people into the various groups. The chi Square table would expand as needed and then we need a sufficient number of observations in each cell.

An analysis of the data would be needed to determine if a player had a significant, real, or "giant" advantage. Most of these types of things are subtle and I suspect there are some "real" but not "giant" advantages. The really big advantages do not need high quality studies to tease out useful variables.

In general a well defined and conducted sudy could be replicated several times and could lead to some fairly sound conclusions.

In a multivariate study several different variables could be used. Creating continuous level of measurements would make for more powerful studies. I guess it comes down to the amount of effort one is willing to expend.

02-28-2011, 06:53 PM
Yes -- definitions etc are a hurdle.
Another little problem iz that all arguements are circular, hencely there iz no such thing az proof -- but this shoodnt be an impediment to getting some pretty usefull info here -- hell, a dictionary iz just a big circular arguement.

If i define a good power breaker az a player who wins more than say 55% of hiz/her break games, then i am doomed before i hav even started. The smallest circular arguement (not meaning a "math arguement") i might ever get.

In the end i think one would find oneself falling back to what i think i sayd earlyr. One shood simply follow the fortunes of individuals, and rank theze "whos" (rank all players in fakt) -- and, their stats will tell us lots, even if we argue about "hows" & "whys".

02-28-2011, 07:32 PM
There is much more information needed in order to determine anything of significance from these numbers.
Since we aren't just talking about break and run, but winning from the break, a power breaker who is a weak safety player might have a great break, but still not win. And a weaker breaker who knows how to run 3 or 4 and then lock up a safety could have a huge advantage.
There is so much more that goes into winning at pool than the break. Yes, the break is important, but there are 8 other shots that are equally important, and it takes all of them to win.


03-02-2011, 01:57 PM
<div class="ubbcode-block"><div class="ubbcode-header">Originally Posted By: cushioncrawler</div><div class="ubbcode-body">If i define a good power breaker az a player who wins more than say 55% of hiz/her break games, then i am doomed before i hav even started. The smallest circular arguement (not meaning a "math arguement") i might ever get.

mac. </div></div>

The argument need not be circular.

Power breaker is one whose CB is traveling 22 MPH or faster during the break shot.
Non Power breaker is less than 22 MPH

In this way type of break is separated from the dependent variable "who wins."

The MPH used could be set as the median break speed for a group of pro players. Then of course we would assess the speed of the break shot to make sure the "Power Breaker" is identified.

If you wanted to see if there was any significant difference then we would set the break speed defining power breaker as greater than 1.5 SD above the mean and non power breaker as 1.5 SD below the mean. An "extreme groups" analysis could be used to determine if there is any advantage to power breaking.

My bet is as indicated above. Regardless of type of break the person who breaks is more likely to win.

03-02-2011, 04:29 PM
Sounds ok.
Re Breaking = winning --- one of my ealyr points woz that good breaking = very hi win %, and that ordinary averages etc only hide that rather than highlite that.