Attn Justin and other SBR $harps, I need your opinion.

**jgilmartin** · 06-18-12, 04:32 PM

I'm assuming the -106 is the average normal (line including vig)? Is the average no-vig line known/able to be calculated with the info you've recorded?

**Cheme82** · 06-18-12, 04:35 PM

Yes, that is the average line I have played. The no vig line (according to me) has been at least 10 cents more for me to even make a play. So it will be somewhere around -120 or so. I can calculate an approximation of what it was but it won't be exact.

**HUY** · 06-18-12, 04:39 PM

If you want to compare your model to the bookmaker you should include the vig in your odds. A line of -106 implies a probability of .5146 if there is no vig. To compensate for the vig (how much depends on your bookmaker) you will have to reduce the .5146 to something slightly smaller, so I would take the second z-score (the one that is calculated from probability equal to .5).

If you want to see how likely you are to win money from the model you should look at the first z-score.

Either way, your numbers look very good, congratulations.

**Cheme82** · 06-18-12, 04:56 PM

So since my average line was -106 does that mean that the book thought that the plays had less than 50% chance of winning since they usually juice them by at least 10 cents per side?

**Justin7** · 06-18-12, 09:54 PM

When you are betting props, you are probably betting them at a lot of different prices. Can you make a spreadsheet, showing your wagers (with prices) and results?

Prop markets are not quite as efficient... In general, your wager is more likely to move the market than in a major market. Consequently, you need to view the line movements a little more harshly than when betting major markets. If you can include line movements in your spreadsheet (betting win % versus closing win %; e.g. you bet +100 at o.5 winp versus -150 at 0.6 winp) that would help also.

Prop markets are much, much easier to beat than major markets. I would not be surprised if you were able to destroy these. You might do more damage than good by discussing what you are betting -- other people will look closer at them. If you include wagers, you might just include a data, and "Event" (with no explanation of what you bet).

**uva3021** · 06-18-12, 11:12 PM

A z-score comparing yourself to plays you have already made is merely trivial and tells you nothing. What you want to do is find x number of games that you didn't bet on in the past that fit a similar profile and see how well you would have done. Then you can compare

**HUY** · 06-19-12, 10:37 AM

Originally posted by Cheme82

So since my average line was -106 does that mean that the book thought that the plays had less than 50% chance of winning since they usually juice them by at least 10 cents per side?

Yes. Normalize your lines so they have no vig (i.e. you can bet both sides and get exactly your money back) to see what the bookmaker really thought. Keep in mind that the vig is not evenly distributed among selections.

**mathdotcom** · 06-19-12, 10:50 AM

Originally posted by Cheme82

I have been playing hoops props for about 3 months, I have made 424 plays so far. Turned a $400 deposit into over 5k of profit. Mathematically, I'm trying to see if I have just been lucky or if I may have something that works.

Using the Z-score test I calculated it for a 51.45% probability of win (my average line was -106) and got 1.55

I also calculated it for a 50% probability of win and the Z-score then is 2.14 (much better).

My first question is: Which of the 2 Z-scores is correct (if any)?

Also, I want to look at the line movement test to get a better idea of whether this will work going forward. Is the prop market and efficient market so that the line test will be meaningful?

Also, which books would you check for closing lines? Pinnacle, Greek, Bookmaker, etc.? Is there a place where the closing lines on props can be found?

Thanks.

One of the following two things is true:
1) your betting strategy is fukked
2) you calculated your z-score wrong

Most likely #1 since if you have to ask this question, you are probably using a stupid system instead of calculating a fair line and then only betting numbers that are better than that line.

Line movement is totally irrelevant here

**mathdotcom** · 06-19-12, 10:53 AM

Originally posted by Justin7

Prop markets are not quite as efficient... In general, your wager is more likely to move the market than in a major market.

Prop markets are much, much easier to beat than major markets.

So given these statements why do we care about the closing line at all in an inefficient market?

**Justin7** · 06-19-12, 11:02 AM

Originally posted by mathdotcom

So given these statements why do we care about the closing line at all in an inefficient market?

When you are betting so much that you control the market -- as in props, it is a big red flag if the market fades you with any regularity. If it moved against me 20% of the time, I would question my methods on the prop. If a major market moved against me 20%, that is not as significant.

**mathdotcom** · 06-19-12, 11:11 AM

Originally posted by Justin7

When you are betting so much that you control the market -- as in props, it is a big red flag if the market fades you with any regularity. If it moved against me 20% of the time, I would question my methods on the prop. If a major market moved against me 20%, that is not as significant.

??????????????
Did I read that right?

**Ken Ruthky** · 06-19-12, 11:24 AM

Originally posted by Justin7

If a major market moved against me 20%, that is not as significant.

If you bet major markets and lines move against you 20% of the time, wouldnt that be considered a truly great achievement?

**Justin7** · 06-19-12, 11:48 AM

Originally posted by Ken Ruthky

If you bet major markets and lines move against you 20% of the time, wouldnt that be considered a truly great achievement?

If I bet before lineups are announced (or openers early in the week for football) and I see 60% disagreement and 20% agreement, that is a good result. If I saw the same result on a prop, I have the best of it, but I am missing something significant. 80% agreement, 20% neutral is more typical for props.

**Ken Ruthky** · 06-19-12, 12:20 PM

in your opinion, what should a good no-vig closing line beating (percentage-wise) be ?

**Justin7** · 06-19-12, 12:25 PM

Originally posted by Ken Ruthky

in your opinion, what should a good no-vig closing line beating (percentage-wise) be ?

On major markets, >50% is good. On props, it needs to be much higher (if you are betting the limit) since you are moving the market. If 100% of your bets move the line, any movement the other way means someone disagrees with you. Most prop players rarely disagree on an individual prop.

**Cheme82** · 06-19-12, 12:40 PM

Originally posted by mathdotcom

One of the following two things is true:
1) your betting strategy is fukked
2) you calculated your z-score wrong

Most likely #1 since if you have to ask this question, you are probably using a stupid system instead of calculating a fair line and then only betting numbers that are better than that line.

Line movement is totally irrelevant here

This is exactly what I'm doing. So I probably calculated my Z-score wrong. I used my average line to calculate the winning %, that's what gave me the 1.55 which translates to about 1 in 17 chances that I have been just getting lucky.

The second value was calculated using 50% win probability. That gave me a Z-score of 2.14 which translates to about 1 in 62 chances that it's just luck.

Finally I just calculated it by normalizing my lines (thanks HUY). That gives me a Z-score of 2.53 or about a 1 in 175 that this has been just pure luck.

**mathdotcom** · 06-19-12, 01:36 PM

Now cheme tell us about how much you decided to bet on each play.

If you turned $400 into $5000 using around 5% of your BR each time I think your z-score would be a lot bigger than 2 son

**Cheme82** · 06-19-12, 02:42 PM

I was too lazy to do kelly. To be honest with you I wasn't even planning on playing this year. I had just started a new job, grad school, kids sports, etc. I didn't have the time I thought I needed. Bovada sent me a 25% deposit bonus up to $100 so that's why I sent the $400. Then I focused on props only because I had played them last year and had good results..

Back to your question. I would calculate the fair price of the prop and then would bet anything that was off by at least 10 cents. I would play to win (or risk if +odds) 5% of my bankroll for a 10 cent difference and increase 5% for every 5 additional cents I believed the line was off. At the beginning that was very risky since a lot of times the whole bankroll was at stake, but as the bankroll grew it was safer. Eventually I just settled to $100 on a line off by 10 cents and $50 for every additional 5 cents the line was off because I have $250 limits on props now (don't know if it's because of the playoffs or not), and most of the plays i find are somewhere between 10 and 20 cents off.

**mathdotcom** · 06-19-12, 02:56 PM

Originally posted by Cheme82

I was too lazy to do kelly. To be honest with you I wasn't even planning on playing this year. I had just started a new job, grad school, kids sports, etc. I didn't have the time I thought I needed. Bovada sent me a 25% deposit bonus up to $100 so that's why I sent the $400. Then I focused on props only because I had played them last year and had good results..

Back to your question. I would calculate the fair price of the prop and then would bet anything that was off by at least 10 cents. I would play to win (or risk if +odds) 5% of my bankroll for a 10 cent difference and increase 5% for every 5 additional cents I believed the line was off. At the beginning that was very risky since a lot of times the whole bankroll was at stake, but as the bankroll grew it was safer. Eventually I just settled to $100 on a line off by 10 cents and $50 for every additional 5 cents the line was off because I have $250 limits on props now (don't know if it's because of the playoffs or not), and most of the plays i find are somewhere between 10 and 20 cents off.

So sometimes the line was off by 100 cents, you bet your entire bankroll and never lost?

**Cheme82** · 06-19-12, 03:22 PM

Originally posted by mathdotcom

So sometimes the line was off by 100 cents, you bet your entire bankroll and never lost?

Never found a line off by that much. And even if I had, the limits wouldn't allow me to bet the whole bankroll in 1 play (and even if there was no limit, I wouldn't have done it). I meant at the beginning on some days I had my whole bankroll was in play (shared among several plays).

**MonkeyF0cker** · 06-19-12, 03:30 PM

Originally posted by mathdotcom

Now cheme tell us about how much you decided to bet on each play.

If you turned $400 into $5000 using around 5% of your BR each time I think your z-score would be a lot bigger than 2 son

Something's definitely amiss.

1.55 = [X - 0.5145(424)]/SQRT(424*0.5145*(1-0.5145))

Solve for X.

X = 234.0995149

Win% = 55.1886792%

Did you have 234 wins out of 424, Cheme?

**Cheme82** · 06-19-12, 03:36 PM

Yup, 234-178-12

All picks are posted on my thread: http://forum.sbrforum.com/nba-basket...all-props.html

**MonkeyF0cker** · 06-19-12, 03:49 PM

Originally posted by Justin7

When you are betting so much that you control the market -- as in props, it is a big red flag if the market fades you with any regularity. If it moved against me 20% of the time, I would question my methods on the prop. If a major market moved against me 20%, that is not as significant.

Jesus Christ.

**MonkeyF0cker** · 06-19-12, 04:02 PM

Originally posted by Cheme82

Yup, 234-178-12

All picks are posted on my thread: http://forum.sbrforum.com/nba-basket...all-props.html

Ok. I'm in agreement with uva. Your z-score isn't really going to tell you much here as to whether it will be profitable going forward. You're either finding ridiculously huge edges (and running poorly) or overstaking to get the results you've achieved so far in terms of bankroll growth. You're exceeding full Kelly expectation with a win percentage of 55.1886792% by a pretty big margin (with an expected bankroll of $4228 and a median bankroll of $1308 after 424 wagers). My guess is that you're overstaking fairly significantly. I don't have much in the way of useful advice other than to tell you that you should probably lower your wagers as a percentage of bankroll and employ some better risk aversion going forward. You'd need quite a bit more than 1.55 (or 2.14) sigmas to have any great confidence in what's transpired to this point.

**Cheme82** · 06-19-12, 04:17 PM

What about the 2.53 I found after normalizing the lines?

**MonkeyF0cker** · 06-19-12, 04:30 PM

The question that you should be asking yourself is how you derived the method that you're employing. Did you backtest it against any data? Are the results that you're seeing in line with what you had seen in your out of sample testing? If so, then you can have some confidence. But even a z-score of 2.53 yields roughly a 1 in 175 chance that your results are lucky to some degree. That alone would not give me much confidence going forward.

That still doesn't address the issue of overstaking either.

**Cheme82** · 06-19-12, 04:55 PM

A 2.53 isn't good enough for you? What score do you consider good enough? The overstaking part you are right on though, specially in the beginning when I seriously over bet my edge.

I don't know how to back test because I don't know where to get lines on past props.

**Justin7** · 06-19-12, 04:56 PM

You can't use a z-score alone. It is only one thing to consider.

**MonkeyF0cker** · 06-19-12, 05:01 PM

Originally posted by Cheme82

A 2.53 isn't good enough for you? What score do you consider good enough? The overtaking part you are right on though, specially in the beginning when I seriously over bet my edge.

I don't know how to back test because I don't know where to get lines on past props.

If someone hits an 8-team parlay, what would you infer from those picks and extrapolate to their future picks?

How much confidence would you have that they have an advantage going forward?

That is essentially what you're asking.

**Ken Ruthky** · 06-19-12, 05:02 PM

Originally posted by Justin7

You can't use a z-score alone. It is only one thing to consider.

z-score
mean absolute error/root mean squared error
closing line
?

**Juret** · 06-19-12, 05:09 PM

absolute median error
plot the error distribution

**Cheme82** · 06-19-12, 06:26 PM

If I'm going to check closing lines (I have to check right before the prop is taking off the board) which books would you guys consider "sharp" when it comes to prop lines.

**rsigley** · 06-19-12, 07:54 PM

Originally posted by Justin7

When you are betting so much that you control the market -- as in props, it is a big red flag if the market fades you with any regularity. If it moved against me 20% of the time, I would question my methods on the prop. If a major market moved against me 20%, that is not as significant.

major prop offered by pinny for past 2 years

Steam against my picks: 363-400-428(-24.43u)

Me: 915-718-896 (36.18%) 158.35u

End result: Pinny cuts limits in half for everyone

**hutennis** · 06-19-12, 09:13 PM

Originally posted by Cheme82

A 2.53 isn't good enough for you? What score do you consider good enough? The overstaking part you are right on though, specially in the beginning when I seriously over bet my edge.

I don't know how to back test because I don't know where to get lines on past props.

Not knowing how to back test and where to get lines on past props is the least of your problems.
In fact it is not a problem at all. You can get nothing out of it anyway. It is useless.

"A 2.53 isn't good enough for you?" - this what seems to be a problem.
Looks like you dont really understand how laughable this question is.

Think about it.
We need only 175 people to expect one of them to have 2.53 zscore just by good old luck alone.
I dont know how many users are at this forum, but i'm sure enough to have a bunch 2.53 zscores who got there just by being lucky. Nothing else.

If you divide number of people in a world who play this game by 175 you will see that a few armies can be assembled from excited guys who are sure they are on to something when in fact they are nothing but a bunch of lucky dudes.