Betting Models

**Miz** · 07-26-18, 10:23 PM

Hi Silver, you can share an email address on here if you'd like, and I can reach you there

**nopressure** · 07-29-18, 04:28 PM

Hi Miz,

I have noticed that in men's tennis ,books tend to inflate the spread and match money line in favor of the the winner of earlier sets. The fact is that players tend to take a set off if they know that they can come back in the next one(s). I don't have access to live line past history for an accurate back test , but I am pretty confident that I can yield a positive return with a few other filters.This model probably does not apply when top players get ahead of much lower rated ones.
This simple idea can be applied to other sports with the right conditions.

What kind of backing can you provide ?

**Miz** · 07-29-18, 05:03 PM

Hi NoPressure,

thanks for the reply. I think the first step in your idea is to generate a set of quantifiable conditions, mine/scrape the required line/stat data, and backtest it for some reasonably large number of matches. Once that is complete, it can be objectively analyzed and assessed for profitability. If you'd like guidance in how to do this, just ask.

Thanks!

**tsty** · 07-29-18, 09:30 PM

[QUOTE=Miz;27938637]I am curious if anyone has a fully backtested model (or models) that they bet, but they lack a large enough bankroll to see significant and worth-while growth. /QUOTE]

lol

if it was only that easy ey

**tsty** · 07-29-18, 09:31 PM

Originally posted by nopressure

Hi Miz,

I have noticed that in men's tennis ,books tend to inflate the spread and match money line in favor of the the winner of earlier sets. The fact is that players tend to take a set off if they know that they can come back in the next one(s). I don't have access to live line past history for an accurate back test , but I am pretty confident that I can yield a positive return with a few other filters.This model probably does not apply when top players get ahead of much lower rated ones.
This simple idea can be applied to other sports with the right conditions.

What kind of backing can you provide ?

Live Tennis is unbeatable.

**Miz** · 07-30-18, 05:33 AM

[QUOTE=tsty;27948042]

Originally posted by Miz

I am curious if anyone has a fully backtested model (or models) that they bet, but they lack a large enough bankroll to see significant and worth-while growth. /QUOTE]

lol

if it was only that easy ey

lol, I figured hey why not dangle it out there. I was once a struggling artist playing dive bars too.

**oilcountry99** · 07-31-18, 07:29 AM

Does anyone have an MLB model that uses BaseRuns (BsR) developed by David Smyth?

**JacketFan81** · 08-03-18, 07:26 PM

Originally posted by oilcountry99

Does anyone have an MLB model that uses BaseRuns (BsR) developed by David Smyth?

I have 1 that uses both BaseRuns and Runs Created

**oilcountry99** · 08-04-18, 06:47 AM

Originally posted by JacketFan81

I have 1 that uses both BaseRuns and Runs Created

I messaged you

**Hgondorff** · 08-06-18, 01:44 AM

Hi Miz! Yeah, this is a great topic. First time poster, long time user of this site. A lot of 'sharp' players on this board and I thank all who contribute.

I am a trend player. About 4 years ago, I had an idea and started to experiment with it. My goal was to design a betting model that could be completely 'mechanical' and produce a 'positive expection'. Insane, right?

But, as you said in your post, I have learned a lot while doing this. I am still working on this project and feel I have gained a lot of knowledge in the process. I am seeing things more through a mathmatical perspective, situations where all the wishful thinking in the world won't change the math. I have spoken to other players who have tried to do the same thing. You can maybe get ahead of the deviation curve in the beginning, but when the deviation goes against you, your bankroll blows up.

I developed a model that is mathmatically sound, but as you said, the amount of action required vs the low percent of return is insane. You would need a huge bankroll and a dozen 'books' to make it work.

However, I am still working at it. Everytime I feel I have exhausted every angle, a new idea pops in my head. Got one today as a matter of fact.

I used one of my ideas last NHL season and got good results through the midway point. Then is stalled and the deviation became too irratic, and I began to pull back on bets because I was afraid of giving it all back. Tried to plug the same model into MLB and have exploded my account about a half dozen times. So, looking ahead to NHL I came up with a different approach and realized that I could test it against MLB's last 8 weeks.

Just a 'heads up'...I am pretty proficient in Word Excel but I don't know how to write code, or program. Thanks!

**HeeeHAWWWW** · 08-06-18, 04:00 AM

Originally posted by Hgondorff

I have spoken to other players who have tried to do the same thing. You can maybe get ahead of the deviation curve in the beginning, but when the deviation goes against you, your bankroll blows up.

That usually means you're over-staking :-)

I developed a model that is mathmatically sound, but as you said, the amount of action required vs the low percent of return is insane. You would need a huge bankroll and a dozen 'books' to make it work.

Yes, that's fairly normal. If a market is sufficiently soft to allow sizable edges, the limits will be small. If they're large, you're chasing small edges.

**Hgondorff** · 08-06-18, 12:28 PM

Thanks HH! Thanks for the comments. Yeah, 'over-staking' or 'insuffiently bankrolled', I hit the wall. The models I have come up with require the player to 'bet the board'. Doing this with the hope of 'smoothing out' the deviation curve a bit. But, I still get blown up. I've made a small adjustment starting today of using a 'virtual bet' to off-set some of these swings.

Hardest part - capturing units and securing them while being able to slowly increase my unit size over the coarse of the season.

Thanks again!

**danshan11** · 08-07-18, 10:58 AM

when you say deviation I assume you mean past results and some correlation to future results, is that what you mean?

**Hgondorff** · 08-08-18, 12:32 AM

Hey Danshan11

Let me lay out the 'basic' concept of what I am looking at, in the way of a model. I am certain that I am not the first to look at 'dog plays' in this manner. This is just a theory and I will lay it out so others can check the math, find the 'leaks' and pinpoint the fallacy.

For this example I will use MLB. The parameters will not line up perfectly, but it will explain the concept. I am trying to revise something for the NHL season, but am testing this live against MLB since it's the only game in town right now.

Parameters: MLB - 15 games - 3 games in each series - using a 3 step progression - betting the board - dogs only

There are four distribution outcomes for the winner of a 3 game series:

(a) W - W - L
(b) W - L - W
(c) L - W - W
(d) W - W - W

For this example assume that all the games won were favorites on the moneyline, and the losers were dogs. It would look like this:

(a) F - F - D
(b) F - D - F
(c) D - F - F
(d) F - F - F

Dividing 15 games by the 4 typical outcomes would be 15 / 4 = 3.75. For this example I am expecting no more than a 2 loss interval for series 'a', 'b', and 'c'. I am expecting to lose my progression bet on series 'd'.

The Progression

I have 15 separate 3 step progressions - I will refer to them as 'cells' numbered C1 to C15. I want to assimulate randomness as best as possilbe, meaing that I do not want 2 teams playing 3 consecutive games within the same cell. For my live test I just use the rotation schedule. For example, if a series began on a Mondays and there were 10 games, they would be assigned C1 through C10 in order of the rotation. Tuesdays line-up would begin from the top of the schedule at C11. Night games are days games the following day, and so forth.

This means that while the progressions are actively running, on the first game of a series, the Baltimore Orioles could land on any step of an active progression - Step 1, 2 or 3 - depending on which cell is assigned and what step is active within that particular cell.

I am using a 1 - 3 - 7 progression. This allows me to win 1 unit for every step within the progression. A win on Step 1 = 1 unit, Step 2 = 2 units and Step 3 = 3 units. After a win, the progression resets to 1. Three consecutive losses = surrender loss amount.

The Progression looks like this:

Step 1 = Risk 1 Unit
Step 2 = Recovery Amt Step 1 + 2 units
Step 2 = Recovery Amt Step 1 + Step 2 + 3 Units

So, given these parameters, here is where I need help verifying the math among other things. This is just a basic concept that I am testing to see if it holds enough water. There are several specific things I want to plug in around this, but before anything else, the math has to make sense.

Betting Unit = 1.00 Typical Dog Price for this example = +1.20 1.00 / 1.20 = 0.833

I am risking 0.83 cents to win 1.00

Step 1 Risk 0.83 to Win NET 1.00 Lose 0.83

Step 2 (Loss Amt 0.83 + 2.00 = 2.83) * (0.83%) = Risk 2.35 to Win NET 2.00 Lose 2.35

Step 3 (Loss Amt 0.83 + Loss Amt 2.35 + 3.00 = 6.18) * (0.83%) = Risk 5.13 to Win NET 3.00 Lose 5.13

Cummulative Loss for this Cell = (0.83 + 2.35 + 5.13) = 8.31

This is the math that I need to be checked out. If my progression would be successful in this example 75% of the time in situations 'a', 'b', and 'c', and I would surrender 8.31 in situation 'd' then on way of expressing it over fifteen 3 game series could be:

'a' 3.75 + 'b' 3.75 + 'c' 3.75 = 11.25
'd' = 3.75

The 'a', 'b', and 'c' patterns would produce 3 units continually at a rate average of 1 unit per game.

('a' 3.75 + 'b' 3.75 + 'c' 3.75) is 3 * 11.25 = 33.75

A 3 game sweep pattern would lose a cummulative total of 8.31.

('d' 3.75) is 8.31 * 3.75 = 31.16

Win 33.75 - Loss 31.16 = 2.59

Profit for this example is 2.59

Number of games wagered is 45

2.59 / 45 = 0.057

If the math checks out, then this 0.057 could be the 'mean' to measure the deviation. The ratio of favorites to dogs would determine how this number would move up and down. I am simply curious whether it would have any value as far as measuring deviation swings.

In this example I used 3 dogs over 12 outcomes. I subtituted the word 'typical' for expectation because I am aware that my example does not include many variables. For example: A pattern that is W - W - L. The wins in this pattern could have been 'dog' teams.

Thus, D - D - F. It has been said that 'the dogs' hit at a rate of 40%. That means instead of 3 dogs hitting over 12 games played, 4.8 dogs are hitting within 12 games.

I do not know how to formulate the equation to express this in percentages.

Q: 15 Series - 3 Games each. Favorite to Dog ratio is 9 to 6 for Game 1, Game 2, and Game 3. What is the expectation of a 3 game sweep and how many?

One last piece it that a 3 game sweep at: Gm1 +1.75 Gm2 +2.25 Gm3 +1.60 is going to measure out much differently than a sweep pattern at lower value.

Thanks in advance for everyone's help, and if this problem has been posted 1000 times before...I apologize.

**danshan11** · 08-08-18, 11:37 AM

thanks for the explanation but you are believing past events are somehow connected to future results.
if team A wins game 1, 2, and 3 they lose 4 based on the implied probability of game 4 ALONE
if team A wins game 2 only, they lose 4 based on the implied probability of game 4 alone
if team is A is fave in game 1 and 3 and wins game 1 only and is a dog in game 4 they lose 4 based on the implied probability of 4 alone

**danshan11** · 08-08-18, 11:42 AM

Dog in last 3 games straight is in the 4th game for example

7691-9890 (-0.54, 43.7%)

avg line: 127.0 / -140.5

on / against: -$61,526 / -$29,062

ROI: -3.3% / -1.1%

and the implied probability is +127 and they won 43.7% which means there is no edge, the game is based on the current game and has zero to do with the fact that they were a dog in the last 3 games.

**danshan11** · 08-08-18, 11:45 AM

now a favorite in the last 3 games and magic no edge

6-7785 (0.47, 55.5%)

avg line: -139.6 / 126.2

on / against: -$56,840 / -$33,913

they won at 55% and the line was 57% so basically even with the increased margin of faves , no edge

the moral of the story is there is no edge in anything like this

**Hgondorff** · 08-09-18, 12:47 PM

Thanks for viewing my post and crunching the math for me Danshan11!

**Hgondorff** · 08-09-18, 01:06 PM

Throwing out progressions altogether, even by adjusting the unit amount depending on the average dog price on any given day would not make a difference? Lowering the unit amount when the win expectation to price is less and raising it when average price to win expectation is higher? Same result I am assuming. No PE.

**danshan11** · 08-09-18, 01:25 PM

I think the staking adjustments could help you lose less by using what I call fixed profit bet 1.xx faves and .xx dogs you can cut losses but in reality if you are not able to produce an edge the best you can hope for is to lose less

**Hgondorff** · 08-10-18, 01:01 AM

Thanks again, danshan11 for your explanations and keeping me grounded. For me, and for a while now, baseball has become extremely difficult to bet. Not making excuses. I consider myself a fairly conservative player and I need to find an angle I feel comfortable with - these MLB lines barely make sense anymore. I'm still testing stratagies, and I'll be back! Thanks!