Tanko's NCAABB Picks

[Tanko]

Hitting ~60% over the last 2 weeks.

I've been testing out a new method for selecting NCAABB games. I started applying it January 29th but I've only been posting the plays since Feb 7th on my SBR spreadsheet http://tanko.mysbrforum.com/spreadsheet/ It is hitting ~60% since Jan 29th and about 60% since Feb 7th (see on the spreadsheet). It's up about 19 units.

This is a pure statistical method for evaluating games. It does not account for injuries or W/L trends vs the spread, etc... Applying these other factors may generate better results. It utilizes the 4 Factors of Baseketball Success to estimate a game spread, compares that spread to the betting line, and then makes a selection utilizing a 2.5 pt overlay.

You can see it regularly selects the "dogs" (which I'm trying to understand further) but it seems to be working well regardless. It does not account for differences between team's strength of schedule (although this is somewhat baked into the 4 Factors data) so applying it to non-conference games may not be as accruate. Also, it picks A LOT of games (e.g. 32 games on Saturday).

I try to run it in the mornings, using the morning lines. I'll post it for a couple of weeks to see if it continues to perform well. If someone see an error in the spreadsheet let me know. With all the picks, it is a pain keeping the spreadsheet updated accurately. And as usual, use these picks at your own risk.

Here are the selections for the last couple of days.
Good Luck.

[Tanko]

Today's games are not looking too good for this system.

Here's an interesting data point... Since I started using this method for selecting games (for 21 days now), there have been only been 4 days in which it has hit < 50%. 3 of those days have been Sundays (including today) and EVERY Sunday has been a loser.

Some kind of religious Karma???

[A.M.S.]

SMH

[Tanko]

Still at ~59% (+17.3 units) with the bad day yesterday.
Hopefully, today will be a little better.

[Tanko]

Adding Maryland +2.

[Tanko]

Minus ~1 unit yesterday. At +16.
Hopefully, things pick up today.

[wkupike2000]

Very interesting. Thanks for posting and I'm definitely going to follow.

[Tanko]

Adding Kentucky +5.5

[Tanko]

Good day today.... 7-2
Up 4 units. Up to +20.5 overall.
SMH --> Not today.

[Tanko]

Wednesday plays

[wkupike2000]

Curious, a couple of your lines (Duquesne and Georgia Tech) aren't available unless you tease with another game to buy 4 or more points. Others buying points was a must as well which changes the odds. How does your method account for those or any recommendations? Thanks!

[Shutup]

Love dogs. Like
Colorado St
Air Force
Syracuse
Georgetown
What is the system? Might help to see if it makes sense as opposed to it just is winning now

[Tanko]

Curious, a couple of your lines (Duquesne and Georgia Tech) aren't available unless you tease with another game to buy 4 or more points. Others buying points was a must as well which changes the odds. How does your method account for those or any recommendations? Thanks!

I use early lines for games that are posted the night before the games or very early the mornings of the games. The lines for today's games were what was available as of ~ 11 PM CST last night. You can check this by looking at the betting odds history on SBR. Clearly the lines have moved a lot since now the Duquesne and G. Tech games are at +9.5 and +15.5 respectively and would no longer be bets I would make.

My system/method is described in the first post. I use the "4 Factors of Basketball Success" (Google the term) to generate an estimated/calculated point spread for the games. I also factor in a home court advantage into the calculated spread. Then I compare my calculated spread to the available betting spread to see where the differences are. The games that have more than 2.5 pts delta are the ones the system picks.

Here is an example.
For the Duquesne and St. Bonaventure game today, I calculate a spread of -8.9 for St. Bonnie or +8.9 for Duquense.
Actual spread available last night at 11 PM was Duquense +13. The delta is 4.1 (13 - 8.9 = 4.1), which is >>> 2.5 pt overlay so bet on Duquense. Now that the spread is down to +9.5, the differnce to my estimated spread is only 0.6, which is not > than my 2.5 pt overlay so I would pass on the game.

As spreads move throughout the day, prior to game time, I may add bets that meet the 2.5 pt overlay criteria. That is why you may see me add a game or two in the afternoons/evenings sometimes. I don't camp out watching the odds move so I may not always add games as spreads move.

[wkupike2000]

Ah okay, appreciate the detailed insight very much. I read the first page so don't think I glossed over and as I love to learn this stuff. I had somehow missed that the lines could be from late the night before. I now know to avoid those games if I can't get the lines listed. Been watching the last few days and ready to ride the sheet of picks tonight. Thanks again!

[Tanko]

Love dogs. Like
Colorado St
Air Force
Syracuse
Georgetown
What is the system? Might help to see if it makes sense as opposed to it just is winning now

See post #13 for some of the details on how the system generates picks. The only additional information to add to that post is how you convert the "4 Factors for Basketball Success" into an estimated spread.

Here are the 4 Factors with the weight applied to each factor to generate a spread.
There are 4 factors for Offense and the same 4 for Defense.
- Shooting (40%)
- Turnovers (25%)
- Rebounding (20%)
- Free Throws (15%)

This link to The Sporting News provides some insight into how the calcs are done. They used simliar calculations for making tournament picks. The exception to their method is that I do not apply the portion of the analysis they are referring to when it comes to tounament seedings.

http://www.sportingnews.com/ncaa-basketball/news/march-madness-2017-how-to-pick-an-ncaa-tournament-bracket-using-the-four-factors-of-basketball-success-stats/14xacw0ak7g9n1gzblode9oyjp

I can PM you an example of one of these calculations if you would like. I hesitate to post it in the forum since it requires using statistics that are obtained from another site that may not like being mentioned on a sports betting website. I don't want to upset SBR or the statistics web site by accidentally posting something that I shouldn't. If you read the article in the link above, you should be able to figure out where the stats are coming from.


This calculation is not exceptionally difficult especially when you dump the data into a spreadsheet. It is probably similar to something others are doing but they are probably filtering and refining the data even more to come up with picks. PM me if you want specific examples.

[Tanko]

Although the system seems to be hitting well over the last 3-4 weeks, the law of averages always comes around to even things out. Having done this type of analysis on NCAAF and NCAABB for several years, with a variety of "systems" to make picks, I realize it will likely turn south at some point. So apply at your own risk.

[Tanko]

Adding:
Evansville +5.5
Siena +4.5

[Tanko]

Not a great day yesterday. 6-9. Down to +16.5 units

[Tanko]

Thrusday games.
I anticipate adding a few more as the lines move during the day.

[Tanko]

Adding Arkansas St. +2.

[Tanko]

This system in general picks “Dogs”. I questioned if it was doing well over the last few weeks due to a general trend in “Dogs” winning or if the system had something more to it than that.

To answer this question, I trended EVERY game result (not just the ones selected by the system) that I had spreads/results for from 1/29/18 – 2/21/18 (~1100 games). If the “Dog” covered I gave the game a +1. If the “Fav” covered I gave the game a -1. Then I cumulatively added the total over the time period from game 1 to game 1100. The blue line shows the cumulative total. As you can see, for the first ~500 games (through ~ Feb 8^th), the “Dogs” performed better (positive slope of the blue line). Then from game 500 (~Feb 8th) to game 750 (~Feb 14^th) the “Favs” started covering more games (negative slope of blue line).

From game 850 through ~ 1000 it was break even with the “Dogs” and “Favs” splitting who covered (~ flat line during this period). In general, I would expect this trend long-term to be relatively flat.

I then trended the SYSTEM picks. Remember, 90% of the SYSTEM picks are “Dogs”. If the SYSTEM was right, I gave it a +1 and if was incorrect, it was scored a -1. I then cumulatively added up the scores and trended it on top of the previous data (see orange line on the chart).

What you can see from the orange line is that when the “Dogs” are winning (games 0-500), the SYSTEM PICKS obviously win. But, what is more important is that when the “Dogs”, in general start to lose (games 500-750), the SYSTEM performance trends down but at a much lower (if not almost break-even) rate. Between data points 500-750 the blue line shows a drop of almost 62% (from 45 to 17). The system picks performance dropped by 19% (peak of 47 down to 38). Similarly, the last ~ 50 games showed a large drop from ~47 to ~36 on the blue line. Obviously there were a lot of "Favs" that wone duing that short period around games 1150. But, the orange line did not drop off as abruptly or to the same extent in terms of percent drop.

This seems to indicate the system is able to more accurately select the correct “Dogs” that will cover rather than just be randomly following the overall trend of the “Dogs”.

Some may not like to read this type of analysis and/or may not like the system in general. That's OK.
I'm just attempting to share ideas and get better at making picks.
Constructive comments are always welcome.

[Tanko]

Adding:
Long Beach State +4.5
Idaho St +6

[wkupike2000]

Appreciate all the insight Tanko! I like how much thought and effort you put into it.

[Tanko]

Adding Oregon St. +6

[Tanko]

Dropped 1 unit yesterday. At +15 units.

[Tanko]

Today's games:
Brown +5.5
Citadel +8
Western Carolina +10.5
Darmouth +11.5
Chattanooga +16.5

[Tanko]

The results of this system have dropped considerably. Could be an 5 unit drop tonight.
We'll try Saturdays' games and if it continues to trend poorly, we'll shut it down.

[Tanko]

Not a great day yesterday. Down to +10 units. Hopefully, Saturday will be better.
If Saturday turns into a bust, it may be time to put this system to rest.....

[Tanko]

This seems to be doing worse than throwing darts at a board of games. Lost 3 units Saturday.
Time to call it to ripe/rotten to use further.

[Tanko]

Since Feb 21st, the favorites have come on strong, covering 18 more games than the Dogs (122 covers out of 226 games). This helps explain why the system tailed off so badly over the last week. From the updated chart you can see the trend for ALL games shows selecting the Dogs is down by 40% since Feb 21st. But, the SYSTEM picks only dropped by 18%, again indicating the system seems to be able select the "better" Dogs than would be selected by random chance.

Since most of the public bets generally tend to be towards the Favs, selecting the Dogs should provide a slight improvement over random selections. One of the factors I'm adding to the system selection is % of Public Bets. The more public betting there is on a game, in general, the more the Dogs should gain an advantage. Why, if public betting on the Fav is 70%+, this should have helped drive the spread to help the Dogs more. Actually seeing the number of bets would be even more beneficial since you would like to see >5k bets on a particular game so that you are seeing A LOT of public bets to help drive the spread in the Dogs favor. Of course there is always a catch. Only a few large bets from Sharps can mess this idea up since they can easily swing lines 1-2 pts and I can't always see when those instances occur while monitoring % of bets.

Finally, we're running out of 2018 regular season games so there may not be enough data to collect to obtain a general trend before wrapping it up.