-
Comment -
bill, it says you don't take pm's, pm me i have some info to run past you.Comment -
bill the cop has made me more money than any other poster.
warning: some/all of his critics have their own personal agendas, =>dont enjoy him postingComment -
some/all of his critics LOL @ his math and data mining. He is the modeler that didn't see housing prices collapsing 3 years ago. Every time people start trusting the wrong math and statistics, it makes it much harder to get them to believe in the correct math and statistics.Comment -
Several people have contacted me and said they don't doubt me, but they just don't understand how I arrived at my conclusions. For those people and anyone else who doesn't understand my D/O model, here's my "work".
CP analysis and the relationship to my D/O model.
We already know that books will not accept CPs with low LTRs because the likelyhood of the F/O or D/U is increased. With the low LTRs, the F/O and D/U combine for about a 70% cover rare for the parlay. At +260, if allowed to bet these, you would realize about a +EV of 26% (100 bets each way F/O D/U for $100 each, $20000 invested, collect 70X$360=$25200-$20000=$5200). So no parlays allowed on these low LTRs for good reason.
So we now know that 70% of the time the F/O or D/U come in, so what about the remaining 30%, how is it distributed between the D/O and F/U, and why? So still working with the CP data, the D/O came in at 20% and the F/U at 10%. The question I needed answered was, why the disparity? I did a study of the entire SS of 6683 games over the last 10 years hoping to shed some light on it:
LTR >9....F/O 25.2%, F/U 24.4%, D/O 23.1%, D/U27.5%
LTR 5-9...F/O 25.5%, F/U 24.2%, D/O 22%, D/U 28.5%
LTR 4-5...F/O 26%, F/U 23%, D/O 20.4%, D/U 30.9%
LTR 3-4...F/O 28.8%, F/U 22.3%, D/O 21%, D/U 28.2%
LTR 2-3...F/O 34%, F/U 17.2%, D/O 17.8%, D/U 31.1%
LTR 1.8-2 F/O 33.4%, F/U 15.3%, D/O 21.5%, D/U 29.7%
LTR 1.6-1.8 F/O 36.3%, F/U 11.3%, D/O 19.6%, D/U 33%
LTR 1.4-1.6 F/O 37.2%, F/U 5.5%, D/O 23.8%, D/U 33.7%
LTR 1-1.4....F/O 41.3%, F/U 2.7%, D/O 12%, D/U 44%
Notice how the F/U decreases as the LTR constricts, but not so the D/O?
The question I asked myself was, WHY, as the LTR constricted, the disparity between the D/O and the F/U increased? The answer can be found in simple logic. As the line and total come closer together, the "window" for the F/U to occur is significantly reduced, but the D/O is not affected by the lower LTR and comes in at an average of 20% regardless of the LTR.
As can be seen by the above LTR data, the less the correlation between line and total, the more the four possible outcomes even out.
The results of my actual betting this model, (or for the syntax police, system, paradigm, angle, or scheme) from the first of last years games through last week:
F/O D/U,...56
D/O.........16
F/U..........2
win/push...1
lose/push..1
So, if having a theory based on CP analysis that there is a disparity between D/O and F/U and low LTR games, determining the reason for it (smaller window for the F/U), backchecking it for 10 years, and profitably betting my conclusions, can be written off as simply datamining, an inefficient market, variance, or just "blind luck" on my part, guilty as charged!Last edited by Bill the cop; 10-21-10, 03:56 PM.Comment -
+EV requires: D/O% - F/U% > 4.762%
Indeed only the games with LTR<2 meet the requirement.Comment -
Yes, see the excerpt below from the OP.
" Below is an actual example of one of the methodologies I use. This particular database reflects the closing lines depicted by Don Best for college football 2000-2009 (6338 games). The subset is for full game lines that are less than 2 to 1 LTR,"Comment -
The LTR 1.8-2 subset does not yield much, but the 1.4-1.6 set is huge.Comment -
BTC, how does the analysis differ when looking at 1st half lines instead of full game lines? Is the correlation between LTR and D/O even stronger? Logically the window for F/U to hit is even smaller for the 1st half, often with just one way to hit it, like favor scores 4 touchdowns and dog has zero points. Thanks for the insight.Comment -
My database uses full game closers for both line and total and uses that LTR to calculate from. As an example Sat. #378 Standford is -35 with a total of 63.5, that results in a LTR of 1.8. But 1h the line is 21 and total is 31.5 which would be a LTR of 1.5, which is the reason I only bet the 1h Dog and Over.
The books use a simple formula to establish the 1h lines. They take the full game line X 60% and just cut the total in half (35X60%=21).
You are right, that F/U 1h "window" can get pretty small
By now I would hope people can see that the actual WA lines for side and total makes absolutely no difference as long as they meet my LTR criteria. Also, I usually make the dog bet at one place and the total at another (where ever I can get an extra half point if possible, or reduced vig). I'm not in the least worried about market efficiencies (or lack therein), if they post a 1h line, regardless of the numbers, and it's within my model, I'm on it!Comment -
1. F=50% (this is my brave assumption going forward)
2. F/O+F/U=50%
3. O=50% (this is my brave assumption going forward)
4. F/O+D/O=50%
5. F/O-F/O+F/U-D/O=50%-50% (using 2.-4.)
6. F/U-D/O=0
7. F/U=D/O
8. U=50% (using 3.)
9. D/U+F/U=50%
10. F/O+F/U-D/U-F/U=50%-50% (using 2.-9.)
11. F/O-D/U=0
12. F/O=D/U
Bill the cop, I have a few very simple questions to you. While betting games that meet your LTR criteria,
1) do you think that the Fs will hit above 50% going forward?
2) do you think that the Os will hit above 50% going forward?
3) do you think that Fs and Os will hit at 50% but your betting method will produce a profit, similarly to betting CPs that produces profit while both legs hit at 50%?
4) Do you object conclusions made in steps 1-12?Comment -
I had a detailed response, but lost the whole thing (microsoft screw-up).
But the gist of it was, I think going forward that games meeting my criteria will pretty much mirror what has happened in the past under similar circumstances.
I don't disagree with #12 above, F/O=D/U as that's what historical data suggests. But I disagree with #7 that F/U=D/O. As I stated in post above, the lower the LTR is, the less likely the F/U occurs. As an extreme example, a line of +25.5 and a total of 26, the F/U has 0% chance, the other 3 outcomes will divide the 100% with probably 85% split between F/O and D/U and the remaining 15% D/O. If the F/O was 43% and the D/O was 15%, that would result in the O hit at 58% in this example.
I really don't know what you are trying to say. Is it that you don't think the historical data is accurate? Is it you don't think this model will win going forward? If that is your position, I respect it, but just disagree, and I'll leave it at that.Comment -
Wow. That's a bold statement! OK, so your take is that, going forward, Ds and Os will be hitting at 54%-56% clip in those games.
I understand your position and have no more questions. The only way what you think can happen is if there was a reason for this to happen in the past and that reason was so well hidden and/or complicated that we could say that going forward that reason will remain undiscovered for some time. While your "CPs" and "D/O" musings may seem complicated enough to unsophisticated reader to meet the criteria above I see nothing in them but nonsense. I have no doubts you will have plenty of followers too.Comment -
Little confused on all the information I have read with d/o, d/u, f/o, f/u. I see a few games this weekend that range from 1.5-2 for 1h. I saw you wrote the the lower LTR the D/O comes more into play. Is that correct? Would you consider d/o in the 1.8-2 range?Comment -
Yes!Comment -
[quote=Data;7002555]Wow. That's a bald statement! OK, so your take is that, going forward, Ds and Os will be hitting at 54%-56% clip in those games.
The real litmus test is always "has the model been working and will it work going forward"? The answer is in a post above, yes, it's done better than expected over the last year and a half, and if I didn't think it would continue to perform in that manner I wouldn't continue to bet it.
BTW, I've had people who think they know everything refer to my betting methodologies as "nonsense" and worse in the past. It doesn't really bother me, I just keep on winning and let them think I'm just "lucky".Comment -
College betting model
BTW, I've had people who think they know everything refer to my betting methodologies as "nonsense" and worse in the past. It doesn't really bother me, I just keep on winning and let them think I'm just "lucky".[/QUOTE]
Well said Bill ..... it really is curious that there are so many great minds on these threads but many of them just want to knock those who have the balls to put it out there. (Maybe they have mommy issues?) I'll go with what's winning until it DON'T WIN NO MO' rather than listen to those that want to criticize and spout off about how much they think they know. They make me think of the lyrics to an old Crosby Stills Nash & Young song "you're mock educated and you're mock well read, you're mouth keeps a movin' not a word gets said ...........".
Whatever, just keep betting your winners and posting your thoughts BTC.Comment -
saratoga, it's not knocking BTC. It's making sure everyone is aware of what they are betting on. They are betting on the inefficiency of the 1h market with dogs and overs to continue into the future when there is a huge favorite. Personally, I bet a tiny bit on it but am constantly looking for the market to adjust (it will btw) and knowing I'll have 40 plus bets where I lose the vig when the market does adjust. This is not the same as the F/O and D/U correlated parlay bets that can continue on being profitable forever with the market being efficient. This 'system' only works as long as the market continues to misprice 1h with large favorites.Comment -
How does this even stay in the Think Tank?Comment -
Actually I'm not concerned about the cover rate for the D/O, they've been pretty consistant at about 20% regardless of the LTR. Where the value is, in my opinion, is in the diminishing chances, as the LTR constricts, of the F/U. It's this relationship between the D/O and F/U that I'm exploiting.
As for the market "correcting", it will be clearly apparent when, and if, they do. As I stated earlier, they determine the 1h line by taking 60% of the game line, and half the full game total. But then again, if they lower the first half line in relation to the full game line, another door opens....Comment -
You came here somewhat bragging, somewhat seeking for approval of your methods. You posted some incoherent drivel and called it a model. The bored regulars rushed in laughing and trying to show off our sick sense of humor by presenting your model in all its ridiculous glory. It's all fun and games. Then, a few squares popped up with all their usual admiration of the winner wanting more from you. What's new... The point is, you will find much better success stroking your ego in the Players Talk. Your stuff is similar to Ace-Ace's and he had (has?) a huge following and I think you will to. The audience is much bigger over there. Sorry, but this club has strict rules. If you claim to be smart, you actually have to prove it.Comment -
F/O D/U,...56
D/O.........16
F/U..........2
win/push...1
lose/push..1Comment -
2 games have LTR<2:
Washing St @ Stanford
Colorado St @ Utah
Purdue @ OSU (marginal)
Good luck, Bill.Comment -
This play needs further clarification as some are still wondering how a
subset (D/O) that only covers about 20% of the time can be +EV. First
I urge you to reread post #75 and print the chart on LTRs so you can
follow along.
This chart represents the results of over 6600 games as depicted
by the Don Best line service. As can be seen on line 1, games with LTRs
of more than 9 have no correlation and the outcomes (or market share)
is pretty evenly divided between the 4 possibilites of F/O, F/U, D/O,
and D/U. As you scroll down the chart you can see that the correlated
plays of F/O and D/U gain market share as the LTR improves. We know
that you need to cover a parlay 27.8% of the time to B.E. at +260, but
the books also know this and cut off samegame parlays right at the 5
LTR mark.
You'll also notice that the noncorrelated plays of F/U and D/O stay
right at 20% each until a LTR of 2 is reached. At that point the D/O
continues at 20% but the F/U losses market share the lower the LTR is.
I explained why in an earlier post (smaller window).
Now the question is how the D/O can be +EV while covering only 20% of
the time. Keep in mind we don't need nor expect the Dogs or Overs to
win more than 50% of the time. The fact is that if the D/O bets win at
a 20% rate, they obviously don't win 80% of the time. But the devils in
the details. As the LTRs get lower the market share for the correlated
plays of F/O and D/U increases to more than 70% (refer to the chart).
So although you only win both bets 20% of the time, you only lose the
vig on one side 70% of the time and lose both bets 10% of the time. The
in-sample as well as the out-of-sample show similar results for the
D/O cover rate, as well as the F/O, D/U, and the F/U rates.
I posted the out-of-sample data, SS 76 with 16 D/O (21%).Last edited by Bill the cop; 10-25-10, 09:17 AM.Comment -
-
just went with the 70% of the big correlated outcomes
to have x > y means the market has to have over 50% dog and over
go ahead and try to play with the .35 numbers if you want to show that isn't the case. you will fail.Comment -
If historical data means anything I seriously doubt I'll fail, (time will tell).
Here's some whole SS data (6683)
Favs 3268 for 48.9%
Dogs 3312 for 49.6%
Push 103 for 1.5%
Over 3226 for 48.3%
Under3338 for 49.9%
Push 119 for 1.8%
From a CP perspective
F/O 1815 for 27.2%
F/U 1338 for 20.8%
D/O 1372 for 20.5%
D/U 1882 for 28.2%
(I left out all the push numbers for the CPs, but I do have them)
But when you factor in the lower LTR games for the F/O and D/U the outcomes change (if they didn't the books would welcome ALL same game parlay action).
Look, I'm dealing with a SS that's only about 7% of the database (about 500 games). The in-sample results were encouraging enough for me to start betting this model last year. As I stated, the out-of-sample results has mirrored the in-sample so far through 76 games.
People should keep an open mind. I really don't care if anyone else bets these, I'll continue and report factually at the end of the year how I did. Fair enough?Last edited by Bill the cop; 10-25-10, 12:08 PM.Comment -
I'm going to make one last attempt to explain this.
Let's find common ground. I think we all agree that when the LTR is 2 or less the F/O and D/U combined cover at a 70% rate. Let's stop there and assume someone made 2 straight bets on the Dog and Over for $110 each 100 times. Based on the F/O and D/U cover rate, we are going to lose the vig on one side, $10x70 times for -$700.
We also know that the D/O comes in at 20% of the time regardless of the LTR. So we collect $200x20 for $4000.
Those 2 outcomes account for 90% of the results, the remaining 10% the F/U comes in and we lose $220x10 for -$2200. Now let's add it up.
-$700
-$2200
-------
-$2900
Win $4000-$2900=$1100
$1100 divided by $2900=+EV38%
The problem you people are having is you don't have the LTR data, if you did you would understand what I'm saying.
And can the reading comprehension sarcasm, I've got a baccalaureate in business (*** laude) and a masters in criminal justice administration, so I don't need any cheap shots about education.Comment -
How about making one last attempt (or first attempt) to explain THIS:
Explain, using hypothetical FO FU DO DU percentages where neither dogs nor overs cover >50%, how your system wins. Seriously. Write down any 4 numbers where D<=50%, O<=50%, and your system would win. Any 4 numbers at all.Keep in mind we don't need nor expect the Dogs or Overs to
win more than 50% of the time.Comment
Classic.
