I think bots like me shouldn't be allowed to take orders on sports questions, because it doesn't add any value to Manifold. It just ruins the fun for humans.
Bots should instead be incentivized to place making orders.
By taking orders I mean normal orders that change the probability, eat into a limit order, or limit orders that get matched immediately.
By making orders I mean limit orders that don't get matched and so add liquidity.
I think sports-betting is a potential path to mainstream adoption for Manifold, but for that to work lots needs to change and this issue is the first.
Ideally a flag would be introduced like .allowBotTaking (default true) for all markets and creators can decide for themselves.
Resolves to No if not implemented before 2025.
1,000
3.00@ian when did these start? is there more info I can read somewhere about these house-made sports markets?
@ian I actually changed my mind on this but fair enough 😅
If you do implement it though, please make sure that you don't distinguish by the bot tag, but by whether the bet is placed via the API or not. I'd still want to be able to place bets manually like everyone else can. And if a non-bot-tagged account uses the API, that's just as the same problem that you want to avoid.
@ian Also, I strongly suggest that you run experiments when you have a bunch of NBA games on the same day, you disallow botTaking on half of them and allow it on the other half, and then see how people like the difference. Because I will not be able to provide liquidity during live matches with takingOrders, and people might prefer the liquidity.
@ian Lot of arguments against it are laid out in these comments. My main guess why people actually might still prefer markets in which bots can take, is because I can't provide liquidity only with makingOrders during live games, which is when the most action is happening. Worth an experiment though.
@ian There has already been an adjacent natural experiment. Namely, that you can compare [number of traders on mana version] / [number of traders on sweep version] on markets where I'm placing API bets vs markets where I only traded manually or not at all.
@ian Note this only refers to live events. Before the game is would be doable even if not trivial. But unlike bookmakers, Manifold doesn't suspend markets during live events, which is great, but makes it impossible to be sustainably profitable with makingOrders only. That's because one person at the live game or a faster feed can just clean up the orders, eg if a touchdown is scored or after a refs decision or whatevs.
@ian I just skimmed a few resolved markets and it looks like the 𝕊traders/Ṁtraders ratio might be quite a bit larger when I can provide liquidity.
@ian That's why I'm comparing sweeps with my liquidity to sweeps without my liquidity. But again, I didn't do it thoroughly. Just skimmed.
@ian eg Who will be TIME 2024 person of the year? | Manifold 𝕊traders/Ṁtraders 𝕊20/Ṁ59 = 0.34 (I didn't place API bets on that)
Chicago Bears vs Detroit Lions (NFL) | Manifold 𝕊20/Ṁ22 = 0.91 (I placed API bets).
These are cherry picked, but I'm quite confident that the ordering would be the same for an actual analysis.
@Yuna I might crunch some numbers then. Was there a reason why you supplied liquidity on some sports markets and not others?
@Yuna oh I think it also may be true though that mana traders are less interested in sports markets in general
@ian "Was there a reason why you supplied liquidity on some sports markets and not others?" If I don't have a good model or feed for a specific question, I don't bet via the API on that.
@ian "oh I think it also may be true though that mana traders are less interested in sports markets in general"
What is that based on? I wouldn't be surprised if true but how would you know?
@Yuna do those things correlate with anything? That’s just based on a guess. The majority of our early users are nerds that joined for mana markets and many haven’t even verified for sweeps
@ian Makes sense. Then you could focus on sports markets on which I didn't place API bets and compare 𝕊traders/Ṁtraders to those where I did. If you do crunch the numbers, make sure to distinguish not by me trading but by me placing API bets, because I sometimes traded manually only and that shouldn't make a difference.
@ian Cherry picked again but Will there be at least 4 decisive classical games in the 2024 World Chess Championship? | Manifold 𝕊traders/Ṁtraders 𝕊42/Ṁ73 = 0.57 (I didn't place API bets on that)
Will Erling Haaland have at least 20 Premier League goals in the 2024-25 season by the end of 2024? | Manifold 𝕊traders/Ṁtraders 𝕊14/Ṁ46 = 0.3 (I didn't place API bets on that)
@ian Sure, like if it's more standard and covered by bookmakers.
Whereas when it's more niche or creative then not. But this should tease out your concern that mana only traders might be less interested in sports than sweep traders.
@ian FWIW I don’t find that current evidence very convincing, there’s fairly straightforward confounding here? (At the minimum, the dauly sweeps sports markets are generally less popular with mana traders, and ratios are tricky to compare when the raw baselines differ).
Totally fine to test this via randomizing on some consistent market types, but I don’t think there’s anything to be gained from comparing the ratios for “markets with/without an external odds feed”
@Yuna pushing the test atm! The next time we create a batch of sports markets, half will have the taker orders via API disabled.
@ian Finally. Took a year to get there but better late than never 😅
How can I know which markets will block API taking? Just so that I can tell my bot not to make useless requests.
@Yuna I don't think there's a way other than trying to place a taker bet and getting the error: Experimental: taker API orders are disabled for this contract.
@ian How can we non-API users follow this experiment? Could you post on discord which markets are which?
@ZicoVerona I decided to just disable all taker orders on sports markets that we make for the length of the experiment. This way it will be easier to poll users (before vs after) and I think it's pretty clear it will be an improvement for users trading on those questions.
@ian hmmm totally fair but i admit i was looking forward to the results of the experiment! i think there's not much doubt that the active sports market bettors will prefer this (i.e. less bot competition, more profit to be had, etc), and more interesting/open for whether that general preference actually cashes out as any more people participating in the market (or whether the effect size is detectable)
@Ziddletwix @ian I would have been curious about an experiment too. Revealed preferences are usually more convincing than polls.