Computer Programs That Analyze Games

I don't mind forming up a group or team to build something or tweaking what I have. How should we go about doing this. We all ultimately have the same goal, lets get started. Any suggestions or questions post them or PM

As far as the basics of how mine works, or what I used to calculate the lines and totals NBA example:
TEAM A (Home) VS TEAM B

For TEAM A average points: TEAM B Points allowed per game PLUS randomly (-3 to 3), PLUS TEAM A Points per game PLUS randomly (-3 to 3), + Home Court which I use (-1 to 3), DIVIDE by 2

For TEAM B average points: TEAM A points allowed per game PLUS randomly (-3 to 3), PLUS TEAM B points per game PLUS randomly (-3 to 3), PLUS randomly (-3 to 3), DIVIDE by 2

Simulate those events 1000 times, record or find out how many times either team won and by how much on average. And add together each teams points to find out the average total of each game. In all events of a tie divide the ties by 2 then split evenly between the teams.

There you go, that pretty much sums up what I did. Honestly the hardest part was putting it all down and replicating each game 1000 times. That and making it organized, I hate clutter. Also making excel understand what I was trying to do with the numbers I put in. Then I got fancy with it and added a "dynamic" line and "dynamic" Over/Under. All that does is, type a number in and it searches for how many times that particular line or over/under hit, or projected to hit percentage wise out of the 1000 "simulations" here are some screen shots of that, using tonights LA Lakers and Charlotte game.

The main problem i see here is sample size. There has to be a certain starting point in the season. For NBA it should work ok, but for sports like NFL with much fewer games, there should be different ways. But this is a solid start.

What do you do about injuries? Nuggets with no Lawson tried to use Foye and Fournier for pg duties this changes the whole game.

Injuries for me come second. If I like a team then find a key injury on my side I stay away, however if there is a key injury on the other side I like it more. But yeah what I have doesn't factor in any injuries at all.

I can't think of any way to incorporate injuries, only a fantasy type thing tracking each players average stat and simulating the same type of thing with minutes, of course a lot more work but could definitely be worthwhile once completed.

One way to account for injuries in basketball is to determine a player's value compared to replacement level, and the player that will be starting in front of the injured player's value compared to replacement level. A replacement level player is worth about 2-3 points below average (average being 0 points). You might also try accounting for the pace (Point Guards and coaching are very important factors when it comes to pace).

Hei, JJ. Very intresting. Just have a question. Do you have an easy way to explain how to run simulations in Excel? I am not a hardcore excel user, so any info on how you did it or where to find info would help alot. thx

Originally Posted by JJones07

Well, I created something like that, took about 2 years working by myself using only excel. I did this for MLB,NFL, NHL, NBA I took the data from ESPN imported into excel using data import, forgot exactly step by step. But once I got it in I added a lot of formulas, algorithms, to make it easy than typing each teams stats. Then used those teams stats that were selected to simulate 1,000 games, then from each of those games I got the average total & point differential. Of course I tweaked the simulation formulas to get more accurate results producing Totals and Spreads that were as close as possible to Vegas's lines. After I added a lot of Dynamic stuff, like typing in a spread to either team and project if they should cover that number, the same with the total. Of course I've been using this to help guide me to spot games numbers that were out of whack and bet those games. I never kept track over the years my winning percentage or record, until recently here on SBR. But I can say I have won more than I lost bet number wise, but got really confident on certain games and bet more than usual and lost a few of those so... Anyways here are screen shots of todays NBA games.

Originally Posted by kaalhode

Hei, JJ. Very intresting. Just have a question. Do you have an easy way to explain how to run simulations in Excel? I am not a hardcore excel user, so any info on how you did it or where to find info would help alot. thx

When I was in undergrad we used a stat modeling program - don't recall name anymore - but you essentially put in parameters and it modeled out the results. Remember using it to model out the "best" time for hockey teams to pull the goalie when they were losing.

Anyway - models are out there and not all that expensive from what I remember, but I had a "time-bomb" version of it that stopped working 4 months after I got it (was for one class only) so not sure how much real version would be

For those thinking of baseball - I used to have a model years ago to predict baseball scores. Wasn't super sophisticated - you had to plug in the starting pitcher's stats for each team, then a handful of offensive stats for each team, then the top 5 relievers by innings pitched for each team and it effectively told you roughly how many runs each team would score. It worked, we were up money on it, but was a lot of work to get all those stats updated on a daily basis, especially on days there would be no plays from the model. I would be interested in working with others who know how to easily get data to program that model. These day's I really only bet football because its the sport I think I understand best and don't have time to get all the data as consistently as I needed it, but have thought about it in the past that I should rebuild that model. I had built it probably 10 years ago - I am guessing there is better data sources available now

Originally Posted by paydayray

My system won the NFL $1000 bankroll contest at Wagertracker the last two years in a row and my system works best in MLB. I have been looking for help improving it for the last five years. Once I show them how it works they move to vegas and never return my messages. I have found through experience that discovering something new is a bad thing.

Good things come with discovering something new.

Bad things come with poor judgment of people and not ensuring enough intellectual property protection.

Even if you have people sign an NDA, hassles abound when you file a lawsuit and chase after them.

By then, too many people use your system and move the lines against you.

You can find ways to get help without disclosing system details.

The big problem we've been having is finding an algorithm that will produce consistent numbers we are looking for, every sport of course has a little bit different pattern of numbers.

Originally Posted by kaalhode

Hei, JJ. Very intresting. Just have a question. Do you have an easy way to explain how to run simulations in Excel? I am not a hardcore excel user, so any info on how you did it or where to find info would help alot. thx

I just copy and paste the formula that I created over and over in cells, and keep track of the results. The first cell for keeping track of our simulation I would type in the first formula.

Originally Posted by JJones07

As far as the basics of how mine works, or what I used to calculate the lines and totals NBA example:
TEAM A (Home) VS TEAM B

For TEAM A average points: TEAM B Points allowed per game PLUS randomly (-3 to 3), PLUS TEAM A Points per game PLUS randomly (-3 to 3), + Home Court which I use (-1 to 3), DIVIDE by 2

For TEAM B average points: TEAM A points allowed per game PLUS randomly (-3 to 3), PLUS TEAM B points per game PLUS randomly (-3 to 3), PLUS randomly (-3 to 3), DIVIDE by 2

Each of those formulas put in a different cell, personally I put the home team on the bottom to coordinate with Home vs Away format on a typical website. The same exact formula you put on top, the away team, put it in the cell on the right, (the next one). Then repeat putting the same exact formula in the next cell, and the next cell for example 10 times. Then on the bottom cell, you are going to put the home team formula, under each cell you did for the away team. Again, put that exact same formula in the next one, and the next one matching the same amount you did for the away team, so 10 times.

As you enter the formulas in each cell, the numbers that appear should change each time you edit the next cell or refresh the data. After you put all the formulas in each cell you can refresh the data by pressing "Command and the plus/equal button" on a mac, I believe you can press F9 on windows.

Now we will refer to each set of cells, one on top and the one on the bottom, as a game. We should have 10 all together. Now we have 10 "simulations" if you will. From here I get the average of every away team divided by 10, our current sample size, subtracted by the average of every home team. That result is what I use for what I think the line should be. You might not be able to get a good idea of how the game will end up, because of a small sample size of 10. You can add more "games" to increase your sample size. I personally found that 1000 to be pretty consistent. And by consistent I mean, the average differential of each game every time refreshed doesn't change dramatically.

I hope I explained that clear enough.

thanks Jonesy…the information here now is what I was looking for in the beginning.

Originally Posted by Git Lo

What do you do about injuries? Nuggets with no Lawson tried to use Foye and Fournier for pg duties this changes the whole game.

Originally Posted by JJones07

I can't think of any way to incorporate injuries, only a fantasy type thing tracking each players average stat and simulating the same type of thing with minutes, of course a lot more work but could definitely be worthwhile once completed.

Well, I been messing around with excel again and the process is a lot faster than I thought. Long story short I've been able to to simulate games with players minutes and... check out the difference, using the Miami/Clippers game the other night.

System 1 The original team stat based and
System 2 Player minute stat based.

Obviously the game was played already so I went ahead and used the box score from that night to put in for their minutes. Inserted a simple stat/minute algorithm for each stat plus randomly -5 to 5 for points, -3 to 3 for rebounds as well as assists to simulate each individual stat based on their minuets and well, I'm happy with the results. Obviously it's not perfect, for low minuets occasionally a negative number pops up in the stat column, due to a random negative numbers being added.

Anyways for the next couple days I will be using teams previous box score minuets to use for their next game "simulation" and see if their is an advantage vs the team stat based simulation. Anyways here's screen shots of the 2 systems.

I will be PM'ing the select members who are showing interest in "combining forces".

I have some great resources to help get stared.

Also, Justin7's book, conquering risk, is worth every penny.

Linear Regression is the place to start for making a model. There is a 7-step video series on youtube that walks you through it.

I had a small breakthrough last night with modeling.