Computer/System Monday

[Cubbiefan16]

Ok.. I augmented a couple of variables in the system and tightened up a few others. The main change being a dramatically increased emphasis on possessions and offense/defense dissimilarities. Tested into past games, i fee the system is much more accurate with respect to not only scoring margin but total scores as well. Not quite sure what the average margin is yet, but it is much closer than the old system's. I'm going to post all the scores I have for today, for tracking purposes.


UC Riverside 57 @ Loyola Marymount 48, Total 105

UC Davis 62 @ Citadel 63, Total 125

Tenn Martin 64 @ Middle Tenn 78, Total 142

Pepperdine 51 @ USC 78, Total 129

Cleveland St 60 @ Syracuse 75, Total 135

Youngstown St 65 @ Kent St. 78, Total 143

UL Monroe 50 @ UL-Lafayette 59, Total 109


A lot of these scores run very similar to the lines on them. This adds more confidence to the system for me, as lines have absolutely no bearing at all on my scores. I think this will make fewer plays but with much higher quality when a mismatch arrises. The biggest difference today in spread is Cal Riverside (5.5 point diff) and total is UL Monroe/La-Lafayette (17 point diff).

Let's see how it does.. Good luck tonight guys

[mikkii10]

good luck with the update

[ScreaminPain]

It appears we are running very similar models.....

As with your projections, my numbers have no bearing to the posted lines, they simply reflect the expected pace of the game and the expected efficiency of the offensive and defensive capabilities. Of course, the final wagering decision is a comparison of the projected line verses the posted line. I use a difference 4.8 as a cutoff, which barely got UC Riverside in as a play yesterday. My model has them at a 4.9 difference.....(they won by 7)

My biggest concern is always the disaparity in Strenght of Schedule. I've yet to discover a resonable solution relating Offensive and Defensive efficiency to SOS numbers. As you are aware, a team can run up some spectacular numbers against ridulously easy opponents (especially early in the year) which skew the matchup. I'd be curious how you correlate that feature in your model.....or if you do.

As usual, Good Luck on your action.....

[The Frozen Pond]

3-4 last night on the Totals. Keep tweakin' the System.

[The Frozen Pond]

3-4 on the Sides too.

[Cubbiefan16]

Originally Posted by ScreaminPain

It appears we are running very similar models.....

As with your projections, my numbers have no bearing to the posted lines, they simply reflect the expected pace of the game and the expected efficiency of the offensive and defensive capabilities. Of course, the final wagering decision is a comparison of the projected line verses the posted line. I use a difference 4.8 as a cutoff, which barely got UC Riverside in as a play yesterday. My model has them at a 4.9 difference.....(they won by 7)

My biggest concern is always the disaparity in Strenght of Schedule. I've yet to discover a resonable solution relating Offensive and Defensive efficiency to SOS numbers. As you are aware, a team can run up some spectacular numbers against ridulously easy opponents (especially early in the year) which skew the matchup. I'd be curious how you correlate that feature in your model.....or if you do.

As usual, Good Luck on your action.....

The biggest concern in my system is predicting the correct pace of the game. If that is within a couple possessions of being right, the score is almost always very close to the predicted outcome. I am currently working on adding to that portion of the system to make the predictions more accurate overall. As for strength of schedule, I use two things. I use a team's offensive efficiency and their opponent's defensive efficiency. The numbers for each efficiency rating are based on their offensive totals and their opponent's defensive totals obviously (for both teams). I use the average of all teams combined compared to each team's actual offense/defense efficiency and production (defensive/offensive) in a formula that I created. Thus, their efficiency is calculated against certain defenses (against certain offenses for defenses) and therefore strength of schedule.

Not sure that explained everything clearly...let me know if you have any more questions.

Originally Posted by The Frozen Pond

3-4 last night on the Totals. Keep tweakin' the System.

I'm working on it. Like I said, the outcomes are usually very accurate when possessions are close. There is still work to be done I know. But there are some things I intend to add that will improve it I believe. Thanks for the support..

[mikkii10]

are you going to post some about wednesdays game bud?

[ScreaminPain]

[QUOTE=Cubbiefan16]The biggest concern in my system is predicting the correct pace of the game. If that is within a couple possessions of being right, the score is almost always very close to the predicted outcome. I am currently working on adding to that portion of the system to make the predictions more accurate overall. As for strength of schedule, I use two things. I use a team's offensive efficiency and their opponent's defensive efficiency. The numbers for each efficiency rating are based on their offensive totals and their opponent's defensive totals obviously (for both teams). I use the average of all teams combined compared to each team's actual offense/defense efficiency and production (defensive/offensive) in a formula that I created. Thus, their efficiency is calculated against certain defenses (against certain offenses for defenses) and therefore strength of schedule.

Not sure that explained everything clearly...let me know if you have any more questions.
QUOTE]

You can estimate possessions very well from box score stats by using this formula.

FGA-OR+TO+0.475xFTA

These numbers are easily imported into a spreadsheet using Excel. I probably can't post the site name here, but if you PM me I'll direct you. The only drawback is the numbers are not broken down into Home and Away figures, simply season stats to date.

[Cubbiefan16]

Originally Posted by ScreaminPain

You can estimate possessions very well from box score stats by using this formula.

FGA-OR+TO+0.475xFTA

These numbers are easily imported into a spreadsheet using Excel. I probably can't post the site name here, but if you PM me I'll direct you. The only drawback is the numbers are not broken down into Home and Away figures, simply season stats to date.

I know the exact number of possessions of each team in every game they have played, and obviously their average too. But it's much harder to predict the number of possessions in a game when two teams play each other. For example, when Gonzaga (69.5 possessions/game) played Arizona (62.8 possessions/game): my old system had Gonzaga winning by around 20 with a high score because possessions weren't a big enough factor in the system. The new system had Gonzaga winning 68-67, obviously much closer to the actual score because possessions were factored in and pretty accurately. My system projected 64 possessions; in reality there was 62. When I become more accurate projecting the number of possessions when say, a North Carolina (79.7) plays an Iowa (57.5), the system will be very accurate.


Sorry I haven't been posting plays recently...I've been very busy of late. I am working on the accuracy of possessions as that is the main cause of the system's imprecision in my opinion. Will be back soon when I think I am more accurate..

[ScreaminPain]

Originally Posted by Cubbiefan16

When I become more accurate projecting the number of possessions when say, a North Carolina (79.7) plays an Iowa (57.5), the system will be very accurate.

I am working on the accuracy of possessions as that is the main cause of the system's imprecision in my opinion. Will be back soon when I think I am more accurate..

I understand your delimma. Of course, it would be easy to simply average the possessions for any 2 teams, but that would not be perfectly accurate. In my opinion, the variables in the equation FGA-OR+TO+0.475xFTA that must be addressed in a particular matchup would be the TO's and the OR's. Anticipating those items should determine the pace of the game in question. One team will impose their "will", so to speak, on their opponent and force the tempo through Offensive Rebounds or an inordinate amount of Turnovers......thus affecting possessions. Identifying those teams and situations greatly improves the likelyhood of correctly predicting scores. Of course, at this point one is drifting away from a strictly mathamatical computation and into a subjective evaluation of matchups. I'm not sure which is best?..

In my formulas, I make slight adjustments for teams I feel will create more possessions than usual due to the level of their opponents. This adjustment allows me to anticipate the tempo and more closely predict a score.

That being said, this year I have been off by quite a few on some teams. I'm not certain if I am using incorrect pace figures or what. It's early in the season, so I'm hopeful things will regress to smaller fluctuations as teams get a few games under their belt.

Keep up the good work......

[uva3021]

just use kenpom or bbstate and adjust from those numbers however u see fit

[mikkii10]

cubbiefan any plays for today?