JR Miller's record

[sportshobby]

Come on, Slam, don't be so hard on J.R. More than half the regular posters in this forum have differing, firm opinions. I'd hate to see you run J.R. off just because he's like most of the rest of us.

[irish]

Reading some of these posts,I am trying to differenciate between arogance and frustration.Its been my experience that most loosers are frustrated.Why would any MORAN want to run this guy off,when he is obviously the most knowledgeable.Sincerely Irish.....

[SLAM DUNK]

Gooden, nobody said it was easy. That is why serious sports bettors are working day and night to get better. I can tell you from personal experience the same thing that LOGIC
will tell you, the correct thing to do is bet more on your strong games, if you are able to recognize what a strong game is.(A BIG IF).
It is hard enough for people to make money without some selp appointed guru, who doesn't have a clue telling them to do the wrong thing.

Sportshobby, JR Miller DOESN'T DISCUSS ANYTHING. HE EVADES QUESTIONS AND KEEP THROWING UP LINKS TO HIS WEBSITE. You guys can line up 100 deep saying that JR would not
promote himself, but that is the only thing he is doing. Does he dispute anything I say, no, "GO TO MY WEBSITE", like I am going to screw around with a deck of cards. JR, some of us are beyond the 1960's and do our analysis using computer simulations, not a deck of cards and a calculator, but to each their own.

JR, since you like to use Binomial Distribution in your analsis so much(so do I by the way), realize this. If I have 100 60%
games there is a 93.62% chance that I will win over 52 games and thus make money on them. Thus there is a 93.62% chance that I will do better by betting more on them, than by not betting more on them. Not quite the 30 to 80% BS you are spouting about.

As Chick Hearn would say, the refrigerator door is shut, the lights are out the butter is getting hard and the jello is a JIGGLING.

[SLAM DUNK]

Irish, now I am the Moron and he is more knowledgable. Do you actually believe that? Since he is unable to respond to any of my points(Why would he when a convenient website link is so much easier) or anybody elses, perhaps you would like to take a few of them on, EINSTEIN. Nobody is running him off. If you are going to go on a public forum and make a statement that is WRONG, not an opinion, then you should be fully prepared to back it up, with discussion, not links. Also nobody is going to run him off because he keeps giving himself free promotion. I mentioned it the first time, he claimed promotion was not a motivation, yet he keeps posting them links.

By the way you think arrogance, losing or frustration is the cause for this. You really do not have a clue do you? How about helping everybody understand the game a little better. Does that make sense? I could care less what betting strategy you use. But for those who are interested in improving their performance as bettors, I like to give them insight from personal experience. What they do with it after that is strictly up to them.

[J. R. Miller]

If anyone is interested, the fellow calling himself "Slam Dunk" reaches the correct conclusion given the premises he lays out. Where his logic is faulty is in the premises. Trying to handicap your own handicapping by deciding one bet has a 60% probability and another has a 55% probability is an exercise in futility. His logic is the obvious mark of an amateur. Of course, he won't do the exercise I lay out on my website because I seem to have challenged his "macho." To the rest of you, I promise you will be glad you did the test, I guarantee it will save you money, and at the risk of "promoting" my website, here's where to get the test: http://www.professionalgambler.com/debunking.html

Call it promotion if you want, but it's a lot easier sticking in a hyperlink than re-typing the whole damned article here. Good luck, everyone. - J. R.

[Scott Landau]

With all due respect, Slam, as someone who has read many of your quality posts, I don't think JR is in here on a recruiting mission. I'm sure he's well aware that 99% of BW readers are not in the market for his service. I believe he is just trying to share some of what he's learned through the years, and I don't see a sales pitch here.

[J. R. Miller]

This article will also help: http://www.professionalgambler.com/binomial.html

[SLAM DUNK]

Okay, now you are finally making some sense, and I was right all along. I am right, but since YOU do not have the handicapping expertise to identify a 60% game, you call it an excercise in futility. If that isn't the most self centered viewpoint I have ever read? "I cannot do it so nobody else can either". Well rather than saying we agree to disagree, in this case I say we agree to agree. So for those of you out there who can identify a 60% game, double up the bet, if you cannot, stick with the JR MIllER way.

By the way this amateur, unlike you can actually make the distinction, so one final question, WHO IS THE AMATEUR?

[gooden]

Slam Dunk, what I meant by my comment was explained by Mr. Miller: who can honestly say they can discern between your 60% and 55% benchmarks? I don't know your background, but I don't believe you when you say that you can, just I would not believe Jesus, Allah, OR their prophets.

If Mr. Miller had claimed such a facility, then I would have dismissed him LONG ago.

HOWEVER, I firmly believe that very careful, strategic, and conservative variance has improved my bottomline, but I don't wanna get into that.

[This message has been edited by gooden (edited 07-18-2000).]

[zippy]

[Note: I wrote this post after the 3:20 SLAM DUNK post, so it is not intended to be responsive to anything in this thread that was posted after that.]

sportshobby and irish: I’m not trying to run Mr. Miller off, and I doubt that SLAM DUNK or anyone else is either.

Expressing disagreement with someone is not the same as flaming them or trying to force them to leave. Don’t confuse the informative, substantive posts by someone like SLAM DUNK with the ridiculous infantile badgering by someone like noholdsbarred against reno. I hope you don’t believe I’ve addressed you disrespectfully when I’ve disagreed with you. If I was a little sharper in tone with Mr. Miller, it is only because he barged into the forum with an abusive, cursing, “God’s gift to sports betting” attitude, which I don’t appreciate. His being wrong about the “vig” matter is no big deal in itself, but his being so arrogantly wrong and so completely lacking in humility is what rubs me the wrong way. If you feel that some of our posts against him manifest frustration, I suggest to you that this is more likely attributable to his attitude than to our being “loosers” [sic].

Of course, I agree with SLAM DUNK that Miller has a tendency to be evasive in defending his more unconventional views. (Whether, as greg says, he is simply saying outlandish things to get attention and draw more curious people to his website, I do not know. That's certainly possible, but there are enough substantive criticisms of his views available that I wouldn’t focus primarily on an ad hominem like that.) Certainly he’s said nothing in his recent posts that was at all responsive to the refutations of his position on vig that SLAM DUNK and I and countless others earlier have offered. Thus, as promised, I’ll not post on that matter further, but will instead address Miller’s discussion of betting systems that vary the size of one’s bets based on one’s expectations of winning (the “Kelly” system, et al).

First off, whether you tend to agree or disagree with him on this issue, it is not a bad idea to read his article on his website so that you can have a more thorough understanding of what he’s claiming and why, and you can decide for yourself whether he has successfully defended his position. Whether you want to take the additional time to run through his suggested experiment or not is up to you. I think it can be valuable to do the experiment, but only if you also think critically and logically about it and about what really does or does not follow from it.

I did the experiment, which I’ll now describe. (There’s a fuller description of it in the article, if you can’t follow it from what I say. Also the way I did it might differ slightly from what he lays out. I think any such differences are trivial, but you might find a reason to conclude otherwise.)

I shuffled together two decks of cards, and then turned the cards over. I did fifty cards, and then reshuffled and started over and did a new fifty. The first time through, I turned them over one at a time, and the second time through, I turned them over three and four at a time. (Miller suggests doing it in multiples like that, to represent the fact that in real life you often bet more than one game simultaneously.) Any 7 through King counts as a win; any 2 through 6 counts as a loss, and any Ace counts as a push. So the first card off the top—if it is not an Ace (push)—has about a 58.3% chance of being a win. After that, the percentage goes up or down with each card you turn, depending on the breakdown of the remaining hidden cards.

The betting system I tested was to skip the bet entirely if I had a 53% chance or less at a win. I bet 1 unit at 53.1%-56%. (Technically, I bet 1.1 units to win 1, 2.2 to win 2, etc.) I bet 2 units at 56.1%-59% (thus my first bet off the top was a 2 unit bet). I bet 3 units at 59.1%-62%. My system also had increments of 4 and 5 unit bets, but in this particular experiment, the deck never happened to have a winning expectation that was high enough for those to kick in.

Miller says to compute how much you would win or lose based on a system of varying your bets, to compute what your average bet was, and then to go back and imagine that you had instead bet that average amount flat every time. His contention is that in the long run doing the flat betting should make you as much or more profit, with a lower risk of going broke.

Here were my results, after one hundred “games.”:

Non-bets: 5-6-0, 45.5%.

1-unit bets: 15-7-2, 68.2%, 26.4 units wagered, 7.3 units profit.

2-unit bets: 31-17-7, 64.6%, 121 units wagered, 24.6 units profit.

3-unit bets: 6-3-1, 66.7%, 33 units wagered, 8.1 units profit.

Overall record, excluding non-bets: 52-27-10, 65.8%, 180.4 units wagered, 40 units profit.

Average wager: 2.03 units risked to win 1.85 units.

Result if I had bet 2.03 units flat each time, (again excluding the non-bets): 41.39 units profit.

So, in the end, the results of the two betting systems were virtually identical, which is consistent with the Miller hypothesis. (Actually, Miller also suggests that using a variable wagers betting system will almost always render one broke before one can complete this experiment. The most I was ever down was about 10 units early, so unless one assumes a tiny bank roll, I was never even close to broke.) That my bets scored so highly is obviously a lucky coincidence (or at least would have been lucky if I had done this in real life with real money), but of course trials of this fairly short length—100 games—can easily have results that are above or below the expected win rate by this much or more.

Now, my comments and questions concerning the experiment and how best to interpret it:

1. I felt it was especially important to post the results I got with Miller’s experiment precisely because they are consistent with his hypothesis. I believe in being fair. Just because I think he is clearly wrong on the vig matter, and even though I may end up disagreeing with him on this matter as well, it is not my intent to suppress evidence that is in his favor.

2. The very idea of the experiment itself is questionable, based on the fact that we are dealing with a mathematical rather than an empirical matter. Imagine, for instance, that you asked someone “What are the chances of rolling a ‘7’ with two fair dice?” and he said, “I’m not sure. Let’s get two dice and roll them a bunch of times and find out!” You would know by his response that he has misinterpreted the question as an empirical one.

In fairness to Miller though, empirical experiments like this are not always totally out of place in mathematical matters. They can be useful to demonstrate unexpected mathematical truths to those who can’t or don’t want to understand the math or logic itself. For example, if a person insists that “tails” becomes more likely any time you have flipped at least three “heads” in a row, and if he can’t understand a logical explanation of how previous flips have no bearing on future flips, you might well give him a coin, and tell him to flip it a few hundred times to test his hypothesis.

Obviously, though, such demonstrations can have only pragmatic value as supplements to the mathematics; they can never override the mathematics. On this methodological question, I agree with what I take SLAM DUNK’s position to be. He lays it out mathematically why, in the long run, one would do better by varying the size of one’s bets. Miller’s experiment is interesting, but I don’t believe such empiricism can refute SLAM DUNK’s math. For instance, in my 100 trials, I had somewhat higher than expected win rates for my 2 and 3 unit bets (56.1%-62% expectation), but a vastly higher than expected win rate for my 1 unit bets (53.1%-56% expectation). This seems clearly a fluke, and it more than explains why I would have done at least as well betting a flat amount every time.

Is it more than a fluke? Is there some reason to expect such discrepancies to occur in such a way in the long run to make the flat betting system come out equal or better than the variable betting system? I don’t see any such reason. In any case, an empirical experiment cannot establish one.

3. This is a quibble not so much with the theory behind the experiment, as with the actual nuts and bolts of how to carry it out. As Miller’s description of the experiment specifies, I didn’t decide the amount of my “flat” bets until after I finished the 100 trials, and could calculate what my average bet had been. But, of course, it would not be possible in real life to retroactively decide how much to bet like this. My average bet happened to be 2.03 units to win 1.85, but I couldn’t know that until I was done. Since you wouldn’t know in advance precisely what your average bet would be with a system that allows your bets to vary, you’d have to just estimate it.

Note, however, that even if the experiment is changed in this way, the results of my particular series of trials would have been every bit as favorable to the Miller hypothesis. If I had decided the amount of my “flat” bets in advance, I probably would have just made them wagers of risking 2.2 units to win 2 units, given that the deck starts in that range. Betting 2.2 units per wager instead of 2.03, I would have ended up with 44.6 units profit.

4. In his article, in his criticism of these betting systems that vary one’s bets with the likelihood of victory, Miller states that they are really just variations of progressive betting systems that increase one’s bet after a loss to try to recoup the loss plus turn a small profit, and that they share the same flaws.

This is false. As you know if you read the thread on this subject that Survivor started (“What is wrong with this scenario?”), I disagree with such progressive betting systems, and I provided examples of why they will not outperform other systems in the long run under the same initial conditions. However, those systems and what Miller is criticizing here are relevantly different. Those systems, in effect, commit the fallacy that past outcomes affect future outcomes. Once you realize that your likelihood of winning your next bet is unaffected by how many bets you’ve won or lost leading up to it, you’ll realize why those systems cannot work.

The kind of thing SLAM DUNK is defending, however, is quite different. He is talking about situations, real or hypothetical, where the likelihood of victory does differ from one bet to the next, not because of past wins and losses, but presumably because of various factors discoverable through skilled handicapping. If his kind of system is flawed, it is certainly not for the same reason that the progressive betting systems are flawed. Miller is lumping together different betting systems that are not relevantly similar.

Overall conclusion: Especially for the reasons stated in my point #2, I cannot put more than a very small amount of weight on Miller’s experiment. It is interesting that when I ran the experiment, the results turned out to be consistent—at least in some respects—with his predictions, but consistency is not the same as proof. I remain open-minded on the matter, but my position at this time is to agree with SLAM DUNK and to believe that in the long run, a skilled handicapper will do better if he varies his bets at least somewhat with the likelihood of victory.

Of course, don’t take my word for it, or Miller’s or SLAM DUNK’s or anyone else’s for that matter. Read all sides, think it all through mathematically and logically, and come to your own conclusions.

[SLAM DUNK]

Gooden, I do not recall asking you to believe me. I could honestly care less. All I have ever said is that the sportsbettor should use the strategy that works best for his individual capping ability. JR Miller claims that it is best to use only his flat bet strategy, no matter what the circumstances. I have gone on to illustrate why the flat bet strategy is not the most appropriate in some instances, and the only response I get is
that I am an amateur because I am capable of doing something that JR MILLER isn't. But the issue isn't whether I can do it, the issue is whether the people trying to maximize individual profits can do it. They are the ones that know their own personal abilities and will make the decision that is BEST FOR THEM. But if you want further information here is a link to MY WEBSITE.


JUST JOKING!

[wintermute]

zippy

Beautiful post - I may have to move you ahead of Reno as the most talented writer in this forum.

I too tried JR's test although on a much vaster scale.

I determined that the optimum Kelly value for wagering in JR's game is 0.125 of capital. I ran 1000 Kelly simulations starting with a capital of 10000 dollars ( 50 wagers per simulation ). The results were as follows

Made money 658 times, lost 342 times, average capital at end of simulation 18781 dollars.

Using the average wager size from the Kelly runs, I reran the simulations with flat bets. Results were 770 wins, 230 losses, average finishing capital 23401 dollars.

Looks good for JR until you realize that the flat bet simulations are biased. If Kelly does well then the average wager is high and if Kelly does poorly the average wager is low. So you bet more with flat betting when you know you're going to win.

If you use the average wager from all 1000 Kelly simulations then the results for flat betting are 748 wins, 252 losses, average finishing capital 18105 dollars. As noted in zippy's post - quite similar returns.

There's a fly in the ointment however. JR will get wiped out 145 times. Kelly will never get wiped out.

'mute

More to follow - have to eat supper.

[zippy]

Thank you for your kind compliment, wintermute. I value your opinion, as your posts consistently manifest that you are one of the more intelligent, articulate Bettorsworld participants.

I read with great interest your account of your tests of the Miller hypothesis, and I look forward to follow-up details from you. I'm puzzled why in your simulations, the flat bet strategy went broke so often and the Kelly strategy never did. I would have thought that the flat bet strategy would be the less volatile of the two.

Actually, it appears from some of the recent posts that the disagreement here is really over whether a skilled handicapper can reasonably ascertain his expected winning percentage at different incremental levels. (See posts by gooden and J.R. Miller.) This changes the whole nature of the debate dramatically.

I find the skepticism of gooden and Miller on this point rather peculiar. Certainly I can understand someone claiming that all sports handicapping is illusory, that basically you'd do just as well flipping coins. At least that's a consistent position. But it seems like Miller is saying that "professionals" can recognize when a given play has a winning expectation of more than 52.38% and thus is profitably bettable, but that it is a laughable sign of amateurishness to think that one could similarly discern which potential wagers cross the 55% line or 60% line or any other. Why?

If indeed this is all that Mr. Miller is saying, then his point is not false, but trivial. As SLAM DUNK recognizes, really they are just "agreeing to agree." If one really can only discern when a bet is favorable but not to what degree it is favorable, then varying one's bets would indeed be pointless. On the other hand, if one can "handicap one's own handicapping" as Miller calls it, and discern finer gradations of one's likelihood of victory, then varying one's bets presumably would be justified, as Miller now seemingly concedes.

I suspect, though, that Mr. Miller is backtracking here, that really he was not originally assuming that qualification to his defense of the flat bet system. I'll cite two reasons for this. Number one, in his website article, he uses an analogy with card counting in blackjack, and concludes that flat betting is preferable there as well. But clearly in blackjack, one can do precisely what he is denying one can do with sports handicapping, namely, know when the break down of the remaining cards gives one approximately a 51% chance of winning, a 53% chance, a 55% chance, etc. Number two, his very suggested experiment itself is rendered disanalogous by this qualification, for in the experiment, just like in blackjack, one can indeed know the likelihood of winning the next bet with more precision than just above or below 52.38%.

So it certainly would appear that his original assertion was that the flat bet system is as good or better than a variable bet system, even when you can make justified assessments of your likelihood of winning each upcoming bet. His very examples and analogies are of precisely such cases.

Is he now simply backtracking, and seeking to save his assertion by trivializing it?

[mls]

First, I agree that zippy is quite a talented writer. Indeed my first thought concerning him is why he was wasting his time on this forum ... but nice to have him.

While slam seems alittle coarse, I agree that you can distinguish some good bets from better bets. In particular in college basketball where you have 5000 games/year.

If "the line" is 3 points off "my line" I know that in a 1000 bets I expect to win 55%.

But if my line is 7 points off I fully expect to win 60% of those games.

Those figures are accurate for literally thousands of bets over the past 15 years.

And I always bet 2 units on those 60%'ers.

[AussieVamp2]

Well, by definition a Kelly betting system can't go broke, but it can be close enough that practically it doesn't matter, and you couldn't buy a packet of chewing gum with what was left