Circles Off - Ep 152 Telemachus Part 2

Podcast Release Date:
Podcast Link:
hagrinBPR Review Date: 2024-05-07 19:42:33
Rating: 8.25

Notes:
First episode was episode 48.

1:50 - Betting Update: baseball modeling / SGP, wrote an article for Unabated on 9th inning ML flaw. Asks for 15 minutes of Rufus' time as compensation. Ends up joining Rufus' group. Rufus' group 4 people - Rufus, Tom, Tele, silent partner, betting NCAAM & golf.

5:30 - Betting Group Reality vs Expectation: "There's still a number of judgement calls where we don't know the right answer." Love this. "Degree to which your network still matters." Brings up Rufus doing pods & Unabated to maintain the strength of the network. Good point although I don't think these two are remotely the same, but I agree with the overall point, the exposure helps maintain and strengthen the network. 3rd - "degree you can still be irrational about long term results." Amazing point.

11:10 - Model Changes: "Models aren't stagnant. We're tweaking the model multiple times a week." Interesting point then mentions the potential for bugs. Brings up making that many changes introduces uncertainty into "did we do something wrong" and I could absolutely see this breeding uncertainty/paranoia. (From a software perspective, I guess I would wonder how much unit testing they're doing if you're pushing that many changes into "production".)

12:50 - Operating at Scale: "We don't bet overnights." "Trading ROI for volume." Brings up measuring success mainly by ROI where we should be accounting for volume/$s earned. Hard agree - just made this point re: crypto, liquidity & memecoins the other day. "Every morning we wake up we identify our bets 1) what do we hit when 2) who are we going to move through 3) when are we going to move 4) how we're going to attack games that we think will move fast." Have to say, point 4 is pretty interesting to me, not something I do a lot of if at all. Would be curious to come up with a criteria for identifying events that might move faster than others. "9% ROI vs 4.5% ROI but 120x more volume."

18:00 - NCAAM Grind: 1st season did terribly till Jan, then finished 4+%. 2nd season started hot, ended similar ROI, felt better at the end of the 1st season. Likes modeling, likes market dynamics, hates actually betting. (Join the club.) Have tried to lay out schedules, but events within the small group require flexibility. Tele on NCAAM, Rufus on golf/football, Tom on trading, but everyone crosses over. 25:30 - still bet NCAAF, do the Super Bowl, some light NFL.

26:20 - xROI: "The goal is 5%, shoot for 55% around there, bigger focus on volume at lower ROIs." "If we're actually getting 5% ROI, we might not be betting enough or we're participating in too small of market." 29:10 - "Sometimes the vacation we need is a new interesting problem." (Join the club.)

31:30 - Super Bowl Prop Story: Noah Gray u13.5 receiving yards (casino had made a mistake and set it at -1300 instead of -130 and they never fixed it. Tom bet it for almost 5 hours straight (50k). Lost on the last drive of the game.

35:40 - Tracking "OT Flips": Brings up Rufus' work that he published that I have linked to before on xROI which is a great article. Brings up ex of tracking bet on something like u140, regulation ends 122, loses in OT. Was "a good dopamine hit" to see that tracking (very interesting human element). Then we follow up some interesting, useful content with several minutes of "least common denominator" content where we claim things that aren't true like no one addressing they are the recipient of positive variance. Not only is this not true, I've literally had positive shootout variance chats with someone they both know. Save this type of stuff for when you don't have a very good guest on, basically derailed Tele in the middle of a 2 part question. So frustrating.

41:00 - Modeling OT: Hypothetical - ask Tele to model a random 5 minute basketball period. Obv tell him the game state. "I would want to know if it was OT or not, but I don't think it's necessarily explanatory." Some things observable, some things not. "More impact from foul trouble, more fatigue." Coach tendencies, pressure, ref behavior, etc. possibly not observable and "as long as I can convince myself I am not overfitting" that there is something predictive there. Ultimately, OT probably not a different "game".

44:05 - Knowing Group's Action: "Doesn't bother him at all." 1) "When you put a bet down on a market, need to know the impact that action is going to have." Gives good example of looking at your outs, betting some at 10am, the rest at 2pm and determining if this is going to work only if your earlier bets are "truly quiet". "This is a Moving Problem - outs that were quiet may not stay quiet especially as they start winning." 2) Tendencies - "We don't want insiders in the industry we respect getting too much inside of our model for things that will be continuous throughout the season." 3) Experimenting - Due to the daily nature of basketball, can test daily. "Send to these partners first today, see what happens." (very cool split testing). "I genuinely don't want someone getting our stuff at the same prices." "Small enough industry, enough tentacles everywhere that we can see when things move more than they should." (Think people underrate exactly how small the industry really is, I've found its pretty small and really strong OpSec is required at scale.)

51:30 - Interesting Betting Concepts: "Venn diagram of interesting things vs what willing to talk about continues to shrink." "We could debate the EMH." (Sidenote - please don't, Ed Thorp has already given a very sufficient answer.) There's some interesting stuff in non-traditional correlated parlay markets. 2nd - "Your information isn't worth what it used to be worth, sorry about that" - love this quote. "Reverse crossing" - buying people out of the market, might help preserve the ecosystem. Ex - Guy betting 2k when they are betting 20k, the games they agreed on were moving, so they were crossing on the similar bets to keep him out of the market and off his noisy outs. On the disagreements, 2k guy told to bet into the market at the noisy outs.

58:30 - +EV/-EV: +EV, Join a book group. Johnny & Rob "not big book guys". Quelle Surprise, #shockSTUN.