is your bot slow because it gets data slowly (or from a slow source) or is it slow because it executes slowly?

@draaglom depends where i run it. if i run it in london, the polymarket london servers take ~80ms to get the sports info from it happening in the US and then i send the bet request to the manifold server in mid US, another ~80ms, so roughly 160ms late, with each of data and execution being as slow.
if i run it in mid US, it gets data after the event happens, gets settled in the london polymarket servers (~80ms), and updates me in mid US), and then i can execute the trade on manifold in ~5ms.
with polymarket in london servers it's pretty much hopeless

i won't, i basically need a kalshi api key and I don't have one

@Bayesian hmm, why don’t you just build your own bot straight from the primary source?

@bens what primary source

@Bayesian can’t you get live sports score data from other APIs?

@Bayesian that’s true, it’s possible that Polymarket and kalshi odds are not well calibrated as we thought and you can exploit the inefficiencies. I doubt it though since the SIG folks have been working on this stuff for a long time.

@bens that's very sparse info that is hard for me to interpret, and probably still slower than the prediction markets. it is not even close to as efficient, the prediction markets and the sports gambling industry as a whole has millions and millions worth of forecasting technology in place, that is reflected into the PM market prices with short millisecond latencies

you can beat it if you attend the game probably, but not (realistically, imo) by looking at the scoring apis

@Bayesian I checked betting markets price movements while attending an nfl game, and you basically can’t beat it, the latency is negligible by human standards.

@Bayesian I mean, I didn't say it would be easy haha

update: i got some stuff working preliminarily but it's barely an improvement compared to the default. there's something i think would gain a lottttt of speed but I don't have access to it. it is devastating

@Bayesian honestly you can probably boot the bot anyways for recent nfl playoffs, me and cactus been hitting inventory limit insanely fast pretty much pregame and there’s so much EV with not enough takers.

@Mochi so..... you'd need a loan then.

@Bayesian inventory limit as a risk management measure, not in the sense that we run out of investable balance

@Cactus then i can just cover the risk / provide insurance? you capture some of the premium from the trade if we make sufficiently sure that it was +ev with appropriate margins, I take on the outcome variance risk and take a cut

@Bayesian I ll take stakings if you are offering

that would be a good arrangement

@Bayesian I've seen sometimes >10% deviations thanks to certain whales

this is a great development. We gotta work out the details but i'm optimistic. For deviations above 8% where the source you are copying has thousands of $ in liquidity, you can fill it entirely from now on and i'll take a 50/50 EV split I guess? we could also do a profit split but i'm a bit concerned about the risks that have more to do with your bot being exploited / making mistakes and being drained

@Bayesian tbh you can also use your bot (or even manually!) to take the 8% deviations at times they are there for hours because not enough takers.

@Bayesian yea the simplest solution may honestly be for you to run your bot at conservative required edge settings to provide liquidity for us to unload excess shares to you

yea, just set your edge setting at 8% and you can take on a large chunk of the risks. Cleaner method for us all share the risks

FYI NFL whales usually bet for the favorite which causes bots to take large underdog positions more often than not. This increases return skewness a lot

hmm that would probably work but my issue with my bot currently is that i don't trust it to not be catastrophically bad in some high vol cases.. hmmm

it feels like a less efficient solution, sometimes (often) my bot will be down, sometimes there will be huge amounts of mana to be made with less than 8% edges, etc. hmmmm

@Bayesian it’s kinda hard to be catastrophically wrong in these scenarios when the edge is greater than 10 for hours sitting there. You can just do staggered inventory management to not take on too much risk at once.

I still think if you notice a 8%+ edge and you take it manually you can take on any risk under 100k and i'll cover it and take a 75/25 expected profit split. if you do just let me know in the moment, don't need confirmation or wtv

@Mochi yeah if it's pre-game i don't think i would ever go catastrophically wrong. it's just midgame in the last minutes or wtv when price moves 30% in seconds, if a bot is buying it to 60% and my local thing thinks the market price is still supposed to be 30%, i can be eaten up a lot, very bad

@Bayesian you can also consider adding safety measures like redundant REST price checks and timestamp cross reference between the two prices and which may slow the bot down even further but you'd still take the big edges safely

@Cactus yeah but then it would be SLOW

@Bayesian it’s better than manually taking the orders still imo

@Mochi msn games A deviation of around 8% can sometimes remain unchanged for an hour due to insufficient takers.