Superbowl Rams vs Bengals. Dont overthink it.

It very well might not change now.

WoW. Nice stuff. So many questions.

How do you simulate games?

And how do you decide the standard deviation?

And where do these percentages come from?

Is there any way I can chat with you online?

Good to see ya Van.  

Rams should handle this. 

GL pal 

Interesting discussions as usual in a Vanzack thread 

However....with 500,000+ members, even nearing 1% is 5,000 solid guys....kinda eye opening

I'm still debating this one. If this wasnt the SB, I would probably pass. My head says Rams and my gut says Bengals....sometimes one can override the other and compromises can be made....My head is telling me the Rams should win by 3 tds like you predicted.....but my gut chimes in and whispers...."Should?" .....and then goes back about its day. My brain gets defensive and says "Well yeah. They have the better all around team and defense for sure." My gut says "So why is the line only 4 and who is coaching the Rams again?".....This does not make my brain happy.....

@NutinButtLove

I think I saw the last 12 dogs to cover won outright?

I might be wrong about that.

Posts like this are why I continue to post.

Thanks.

@Charon84

 Go read posts #54 and #55 in this thread - I talk about my software and overall betting routine.

If you would like to continue the conversation - I am on twitter and my email is coversvanzack@gmail.com

@THEMUGG

 @searchwarrant

 @Digitalkarma

 @TRAIN69

All of the old timers in my thread.  It is an honor.

GL guys.

(reposting because of request)

Basically - how I decide what to wager and how to stake has a subjective component (my gut), and an objective component (stats analysis through software, and staking through software). 

The gut part is easy.  Every week - before I see any lines - I do an exercise where I sit down with all NFL and NCAAF games - I have 2 columns on a spreadsheet - and I simply guess what I think A. the actual betting line will be (predictive of betting patterns) and in the second column B. what the line should be based on my gut predictive results.  Sometimes A and B are 20 points apart - but that is the whole point of the exercise.

Now for the software / stats side.  In 2006, me and 2 other forum posters got together after participating in a thread about developing some sort of predictive software.  One of the guys is a data engineer, so he really led the way on what should be in there and how to process it.  I am somewhat dangerous in data - but I am not at his level - but we had a collaboration and came up with a model.  We hired people on freelance sites like Fiver to do the work because none of us really knew software.  About a year later - and at a split cost of about 3k each - we had what we wanted and still use it today.

Basically, I have a dashboard where I can give values to importance of many factors.  Things like turnovers, qbs, run game, pass game etc.  Once importance values are set, it pulls stats live off of the internet (the source of stats can be changed fairly easily, and I have done this several times with some help with the advancement of adjusted stats for SOS).  Then I set how many simulations I want to run, and it gives me a distribution that is cut in several different ways.

The single most important result is the median line.  That median line by itself has beaten the final spread at pinnacle for 11 of the 14 years I have used it.  Not by much (best year was 57%), but by enough to give me a start.  It also gives me other results like standard deviation - which is translated in to a variance factor, and final score totals (which as I mentioned in a previous post, are not a focus for me).

(Continued)

I like betting games with high standard deviations.  The way I look at it is....  If the result of the game is within 1 score of the spread - that is a game to avoid (for the most part).  It inevitably happens - but I actually do analysis to try and stay away from 1 score games.  So I either want to get killed in a game, or be on the side who does the killing - the worst result is a 1 score loss.  And standard deviation helps with that - and to determine staking.

The next part is more art than science.  I take my spreadsheet with the 2 columns, take the dashboard 1 pager from my stats analysis and simulations, and I do some voodoo along with some weighted objective values to come up with what makes the cut and what doesnt and a strength value between 1-10 is established for every game. 

My staking is always between 1 and 5 units - rarely 1 and rarely 5.  90% of my bets are between 2 and 4 units - so most of the time I never have more than double my smallest wager on any wager in an NFL season.  I convert the strength value to a staking amount, and that is my target for my wager.

Now.... Things come in to play like line moves, or just expense of getting a line - that make my targets sometimes unachievable.  I also "trade" lines all week by using my predictive numbers to say where a line should move (not results predictive - line movement predictive).  This generates income, but sometimes I get stuck with positions I dont necessarily want because of line moves that go against me - or other reasons.  So I then have to weigh keeping wagers on a side that I have predicted will lose the spread - because the expense of getting out is more than the chance of that team covering.

I try to make all of this as process driven as possible.  I believe in process, and objective over subjective.

Happy to discuss or answer questions if some of this doesnt make sense.

@vanzack

 i need a spreadsheet on all the systems we need to watch closely on this forum.

good laugh on that one. 

Appreciate the thorough reply.  I have been working (cant say where/what) on some predictive modeling, and my biggest challenge is trying to find efficient/effective ways to reduce overall variance driven by large number of inputs and assumptions.  I sometimes get STD Dev so high, that the predictive capability is practically worthless, and so trying to knock down either the number of inputs, or rank them in importance, are a couple of ways to try and come up with a more useful prediction.

From what I understand ("Basically, I have a dashboard where I can give values to importance of many factors.  Things like turnovers, qbs, run game, pass game etc."), you use some type of judgement by weighing the importance of factors, and this appears to be a way of reducing variation in your system.  I found it very interesting that higher Std Dev is your target, but after reading this a couple times, I understand that if you focused on only the low STD Dev games, it would lead to a single play or two that could cause your bet to go from winner to loser, but choosing higher Std Dev games, whereby your model predicts a skew to one side of the median or other, offers much less probability of those 1 or 2 plays biting you in the ass. 

And that last bit factors in to your staking I assume - a high STd Dev + skewed sims favoring one side of the median, probably gets a higher strength rating.  Good to know that you don't purely rely on the model itself.  That is where I always seem to struggle is that addt'l element of intangibles that you and others apply.

Really appreciate your post.  I have always wanted to develop a mathematical model (never have enough time, kids family, life etc.), but now that I am starting to have more, I am just gathering insights to benchmark ways to go about this.

BOL to you Sir! 

@vanzack

 No way. I go to Florida in the winter and the first thing I do when I cross back into Ohio is to stop at Skyline. Nothing better that a five way with onions, beans, and habanero cheese. MMMMMM

@Getty3

 Nothing better than a Skyline 5 way. MMMMMMMMMM!!!!''

@TRAIN69

 THIS!

That's what I have is a mathmatical model. I do exactly what you are saying. I weight different areas and give each area points, then just add up the points and that is a teams Power Rating.

The way I determine which areas to use and how each  areas gets weighted is by looking at which stats have the highest correlation to victory and the highest predictive value looking forward. The way I find this info is off other web sites.

The stat with the highest correlation to victory is also the same stat with the highest predictive value so it's a no-brainer this stat carries the most weight.

To determine how many points each stat gets I breakdown past strong SB winners and try to match the meterics to fit those teams.

Then I branch-out from there to other strong playoff teams and try to find what point-level fits the majority of those teams for each stat I use . This is where it gets a little tricky, you need to continue adjusting each stats value to fit a number of teams and try to get the model to match each of those teams as closely as possible.  

Ohio style is without beans, I heard

You said not to overthink it. What happens if I did? Then what? 

Don't over do think it.

The SB logo is deep Orange

I agree with your logic

The problem is logic doesnt always prevail

Bengals had no business beating the Titans or the Chiefs yet here they are