Notes

Price games with no math match in upcoming and historical

2024.10.16

Change Best Bet history to reflect PA books and same Kelly sizing as what we are betting.

2024.10.15

Kelly to .40

2024.10.9

Bankroll slightly off for both. Can't figure out why.

Mason was -1751.6362388765433, changed to -1661.236239272776

Johnny was -503.8414275242003, changed to -413.4414275242002

2024.9.20

Overwrite all _nv in spreads, alt spreads, totals, alt_totals to take min of 3 methods

25% Kelly

2024.8.28

Now when taking in data, I take min of multi, add, power

No more moneylines

Alts ONLY from Pinnacle

All that LBB shit didn't work

Same set time for everything in run.py. Makes matching by timestamp easier later

The fav standards just lose money. Keep higher 2% ER and 20% Win Prob standards

2024.8.27

Add pinnacle push table for totals; lower threshold of games (4)

Nope: does not work to just use Pinnacle as mode. The average is much, much better

2024.8.26

Start taking add and power of draws, notably of MLS games

Remove hold prob from game history page

2024.4.19

Remove everything but first 5 innings. (1, 3, 7)

2024.4.18

Dramatically reduce push tables by taking min of power/multi/add. Reduces about 50% of probs, so fewer bets should appear but we should be more confident in them.

2024.4.11

add good/bad/par number

2024.4.4

Something weird happened with this gameid:03ec35462bc64c628cc6464389c2b1ea

It was a 5 inning bet. I think I correctly marked the winner and then accidently undid it. But since the payroll was already processed, there was no harm.

2024.4.2

add winner col intelligence

start taking in mlb

2024.3.28

Add h2h to upcoming and history with no tracking table

Remove soccer and rugby because they never offer action

Create single h2h table and start taking in p1,p2,p3,q1,q2,q3,q4 for nhl/nba

2024.3.27

Taking moneyline min now

Adjust LBB to include any ML that qualifies as a BB

But issue w/ LBB chart since if I allow all BB MLs going forward, the past will be changed

2024.3.25

Had to remove power stuff which kept crashing the site.

2024.3.19

Change push table creation to python and not Excel.

Merge old local db to current (breakfifty) so that our push table could continue uninterrupted

2024.3.12

Cancel total_fd in BB

2024.3.11

Adjust total push table to factor in hold percertage; 29% of our number combos now less aggressive

2024.2.29

Did so much seemingly smart math to incorporate hold percentage ... and it was just as predictive as winz.

2024.2.28

add 2.5% kelly threshold to alt bets

Remove nhl from _modal bets

2024.2.26

START USING CASE TO PERFORM CALCS IN SQL

Add live betting page

2024.2.23

Add open and last to bookTrackerMoneylines.

Complete BookTracker___ open and last where available

Change BookTracker__ loop so open/last captured for everything

2024.2.22

Created total_fd. Book count is currently just total books reporting a game.

Added to book_tracker tables w/ open and last number and open and last prob (refers to home/over)

Things to remember about movement:

  • start gap/end gap/book move/modal move
  • there can be different modal moves for the same game_id because everything is measured from when a book first and last appears. So if there are 15 timestamps showing a move of 2 (147 open to 149 close), but one book enters at timestamp 10 (modal of 148), it will only show a modal move of 1.
  • Everything is absolute value.

2024.2.20

DB crash. Upgrade to paid instance.

Start tracking avg_number minus modal_number. For spreads, it's w/r/t/ the home number.

2024.2.16

LBBs and tracking modal book percentage and alt books

2024.2.2

Remove Pointsbet from PA

2024.1.20

Begin official book tracking process and put it in the 4:55am cron

Expand total push table to -8/8

2024.1.17

Kelly to .5

2024.1.16

Soccer spreads are worthless because no american sportsbook seems to price them. Furthermore, there are 'Asian' stakes which could maybe be interesting since it adds complexity, but ultimately irrelevant since there aren't legal books.

I think a general reason that we rarely have soccer op is that American sportsbooks don't have expertise here, so they are unwilling to ever take a stand and diverge from standard market pricing.

Remove Coral from totals and spreads

Adding in tennis moneylines. Not hopeful for the same reason as soccer: the books aren't invested. Maybe if there was an American?

2024.1.12

Bringing in Euroleague hoops

2024.1.6

Include warnings and tracking of multiple modes

Broaden total considerations to -5/+5 from modal to account for some rather large total moves we've recently experienced

Now winsorizing spread push table

2024.1.5

Get scores changed to just 1 day back; doing this because we are now capturing scores again midday

2024.1.3

BetRivers removed from PA

2023.12.22

Add Spreads to bb

2023.12.19

Back live!

Vectorized moneylines

No longer ingest games more than 7 days out

Games can be in totals but not in total tracker because there aren't enough books at the modal

2023.12.13

Everything broken

2023.12.10

NOW USING WINSORIZED

2023.12.8

Lots of thinking re: mean vs. median.

I think I like the idea of keeping push tables at median.I had even considered removing top/bottom values because of how extreme certain outliers were.

2023.12.7

Looking at the totals, I realized that median is preferable to mode as a way of semi-filtering out outliers. Less aggressive. Everything will still say average.

Remember that if there is no ml_vig for a hockey game, it's because we are taking three-way down to two-way.

2023.12.6

Filter totals both for total tracker and best bet to >=MIN_BOOKS and <=CUTOFF_TIME; ESPN BET live

2023.12.4

Totals working

2023.12.2

Stopped betting

2023.11.26

Cleaned all dbs

2023.11.22

Fixed updated commence times so "last" er will be more accurate

2023.11.20

Hockey and Boxing now using three-way to two-way additional to impute lines from books that price draws.

Add 'favorites' to best bets w/ >=40% win prob and >=1% ER

2023.11.16

Kelly back to .5

2023.11.14

If you need old functions, come back to this day. DELETED so much stuff including dbs.

Remove LowVig from NJ

2023.11.10

Always had Betway; now have it legal in US

2023.11.09

Add 5 soccer leagues: Brazil A, Brazil B, Primera División - Chile, France 1, France 2

Kelly size at .25

2023.11.08

Set timestamp everywhere. Sometimes there is a lag in the API and so spreads get "split" between two minutes, which screws up everything that takes the most recent timestamp for calculations.

2023.11.01

Now allowing same game_id in best_bets if any/all home/away/draw exceed our thresholds

Protect against draws. Half the boxing books offer draws

2023.10.28

Removed onexbet and matchbook from hockey because of draw stuff

Protect against draws. Half the boxing books offer draws

2023.10.27

Removed books

  • Keep BetRivers; Remove Unibet, BetParx, Casumo
  • Keep SI Sports; Remove 888 Sports except for Europa
  • Keep Coral; Remove Ladbrooks except for a few leagues
  • Keep Virgin Bet; Remove Live Score

    • 2023.10.26

    Add 3 new soccer leagues.

    2023.10.17

    Slimmed dbs work start to finish. Postgresql at 19.28% before full slimming.

    2023.10.15

    NHL and Boxing live. Few hockey books price draws and we need to exclude them. Change days from in scores to just 1 b/c it was uncessary using 3.

    2023.10.15

    Added Caesars! Start storing boxing and NHL. First linked to Game table.

    2023.10.14

    Must rebuild DB in separate step before populating; switched history to new dbs for state functionality; missing games that weren't in Game db (usually soccer); noticed some incorrect winners (jets/pats) almost certainly because sql queries were written incorrectly by me (as opposed to the API being wrong)

    2023.10.13

    Create "New" Bet dbs w/ associations. Begin to unwind uncessary game info duplication.

    2023.10.12

    No api results for MMA. Have to manually enter.

    2023.10.4

    Add hrs bet db. Set at 6 hrs. $1 ev min. .20 win prob. Surprised only 36 games qualified.

    2023.10.3

    It's not equal black swans? Because my size is static, when a position moves negative, I'll have on too much size (i.e, more than 0), but when it moves more positive, I'll be undersized (i.e., I'm at 1% but Kelly says I should now be at 3%).

    2023.10.3

    Now using 40 sportsbooks. Straight removed TwinSpires and other duplicates from ever appearing in the DB. Added us2,uk, and some eu.

    2023.10.3

    Add bet DB w/ .2,2,5 constraints. Did this because if a game was EV positive but not bettable and became bettable, it would never be recorded in EV and thus never bet.

    2023.9.30

    Sort ML ties based on book ranking (to account for books that aren't bettable)

    2023.9.28

    Added historical games that were missing from EV because of poor code

    Started excluding bad books from calculations; they remain in the DB so we can test in the future if they are still sharing lines

    2023.9.27

    Only carrying bets max of 7 days

    2023.9.25

    Remove live games from best bets

    2023.9.24

    Set minimum book count to 5

    2023.9.24