NBA Algorithm Picks - Day 7 of 10 (27-22 so far))

Welp, the inevitable down day came in yesterday, as the spreads took a beating at 0-4 and the O/U couldn't pick up the slack with a 2-2 split. The difference between going 4-0 on the O/U and the 2-2 result was a total of 4 pts combined (Bucks/Hawks over by 1, Blazers/Nugs over by 3) which definitely is a kick in the balls, but hey, that's sports betting. Sometimes those razor thin margins hurt you, sometimes they help, no point in bitching about it, and a loss by 1 pt is just as valid as a win by 1. At the end of the day, if you continuously make bets with value on your side, the long term record will always be profitable. 

A correction was inevitable in any case. Even the best systems in the world only win at a 60% rate, which meant that our 65% win rate over five days was overperforming at an unsustainable pace. Yesterdays correction brings us down to around a 56% win rate, which is just about the bottom of the expected range of any winning system (between 55-60%) so look for a rebound tonight

To recap:

Overall: 27-22 (12-12 spreads, 15-10 O/U)

Day 1: 8-5 (4-3 spreads, 4-2 O/U)

Day 2: 3-1 (1-1 spreads, 2-0 O/U)

Day 3: 9-5 (4-3 spreads, 5-2 O/U)

Day 4: 5-5 (3-1 spreads, 2-4 O/U)

Day 5: 2-5 (0-4 spreads, 2-2 O/U)

Day 6: TBP

For confirmation of the results so far, click on my name to see my post history and click on the thread for each day. Make note of the timestamps to confirm that all picks were made well before each game and make sure the tally on each day's thread matches the recap above. I can assure you they do, I'm just doing it this way so you don't need to take my word for it.

Here are the lines for tomorrow, as always once the opening lines come out later tonight, I'll post the picks based on that here on this thread:

GS @ BKN------> BKN -1/217

PHI @ UTA------> UTA -3.5/212

SAS @ LAC------> LAC -5/219

Having a down day gives us a good opening to talk in general about one of the more fundamental concepts of sports betting, something that's revelent to any betting system, the relationship between probabilities and sample sizes. Any probability-based event - whether it's a sports event or a coin flip - always has an inherent amount of complete randomness built into the result. Randomness is, by definition, random and simply cannot be accounted for. That's why the sharpest sharps in the world only win at a 60% clip, and why "locks" are a myth that don't actually exist. There's no such thing as a "sure thing" in any probability based event, especially a complicated one with many moving parts such as a sporting event.

Think of a coin flip. You'd rightly consider someone a fool if they said they had "a lock" on any given coin toss (assuming the coin was fair of course), because we all instinctively know that even though the probabilities of any coin flip is 50%, the result of any one toss is completely random. And a coin flip is pretty much the simplest of all probability wagers, so if it's clearly impossible to have "a lock" on a coin flip, think of how much more inconcievable it is to come up with a "lock" on something as complex as the probability of a basketball score, which in reality is actually the sum total of thousands of mini-probability events.

This is why no system can ever be evaluated by the result of one, or even ten, games. Because it's impossible to tell if the results are from the inherent randomness that exists in all probability events, or by actual skillful handicapping. The only way to evaluate a system is over time, with a big enough sample size. Opinions differ on how big a sample is needed before we start digging into the true accuracy of a system, but for sure, a minimum of at least 100 bets is needed before we even begin to assess that. That's why I chose 10 days to run this project over, as it should get us to roughly 100 bets and make for a more accurate assessment of it's accuracy than had I just given out a few games of data.

To easily visualize this, let's go back to the coin flip analogy. We all understand that the probabilities of any given coin flip are 50/50. And we all understand that if our sample size is only 10 flips, it wouldn't be unusual at all to see 7 heads or 7 tails. And we can all imagine someone who played a system of only calling heads MIGHT think they've got a winner on their hands. After all, they're winning at a 70% rate, who can argue with that? But we also all instinctively know that a 70% win rate over, say, 100 coin flips is mathematically impossible. As the sample size gets bigger, the results of an event will always skew towards the probabilities. This Law of Large Numbers is as fundamental to our universe as gravity is, you simply cannot beat it, you can only account for it.

If we increase that sample size to 100 flips, that 70% win rate is simply impossible. At best you might get something like 52/48 (with the most likely outcome to be right at the probabilities of 50/50). Increase it to 1000 and at best you might get 50.5/49.5%. If you had a million flips, maybe your best hope is 50.00000001%/49.9999999999%.

The exact same mathematical laws are at play with sports betting. Obviously, we can't make a million bets before we're satisfied that our system is a winner, there will always be some uncertainty with any system. But for sure, at a minimum, we need 100 plays before we can truly begin to assess accuracy, and probably 200-300 before we can really nail down our true win probabilities.

Only three games tonight, but still got bet signals on 4 out of the possible 6 bets tonight. As the SAS/LAC lines opened up pretty much right where the algo sits on both the spread and O/U, there's no play in that game. As always, since our lines are set before the Vegas ones come out, we always bet the opening lines and the picks I post here are based on those. Almost no movement at all on any of the games, so still lots of value left on the table.

Spreads:

GSW +3.5

PHI +8.5

O/U:

GSW @ BKN: Under 220.5

Phi @ Uta: Under 217

Side note, as this system doesn't bet moneylines, I won't include them in the official tally, but I love GSW to win outright tonight, and at +3.5 odds I'll definitely be putting some of my play-around money on GSW on the moneyline tonight, just for fun.

Your had a line of Brooklyn -1 and the spread is Brooklyn -3.5. This is less than your minimum of a 3 point difference but you recommend betting it? 

@BBGunzz

 I do. In general, I'm a big believer in taking out as much personal opinion or "gut feel" out of betting as possible. There's lots of times the algorithm spits out a line that looks wonky and I think can't be right. When that happens, I'll always dig more into the numbers a bit more, and once in a while there actually is a small mistake in the algorithm, usually just a formula entered incorreclt in one of the cells (it's on an Excel spreadsheet). But the vast majority of the time, the numbers are the numbers and if you trust your system you just grit your teeth and make the play. Oftentimes it ends up winning (and it's those bets that end up making you trust your system and not your gut). I'll always trust the numbers over my gut anytime there's a clear betting signall, which I define as 3 pts. So I'll always play anything 3+ points off, and never play anything 2 pts or less. I do give myself a tiny window for my gut, and that's in the case of anything that lands on 2.5 exactly. It's rare enough that it isn't going to have an effect on my bankroll one way or the other, and more often than not, I don't end up playing the 2.5's either. In this case, I just really like how GS looks when I actually watch the games, and it's hard for me to see them as a 4.5 pt underdog to any team this year.

In addition, I also have a second algorithm that I use to act as a second opinion on certain games. It's essentially the same thing, only it's weighted slightly different, the main algorithm looks back 20 games to try to capture the "true" performance of the team and smooth out any outliers, and the backup looks back 10 games to try to capture any short term trends/hot streaks. I never use the 2nd one as a bet signal, and I would never overrule the main one with it. But I'll see what it says when the main algo is on the cusp, and in this case the backup algo gives me GSW as a 5.5 pt favorite. Now, overall, I trust the main line of BKN as a slight favorite. But in a situation when the line is -3.5, the main algo says is should be -1, the backup says +5.5, and my gut says the same thing, that's usually the kind of setup I'll take a flyer on.

Thanks for posting your picks.