I took this screenshot weeks ago, but never really found the right place to raise it. Now I found this market

I understand the arguments that Yuna is adding accuracy, but I think it's very hard to look at this market and say that Yuna didn't hurt the engagement (for the record, while cricket markets are obviously a bit niche, they generally do fine)

All I see here is massive missed opportunities for humans to engage with the site. This bothers me more than missed opportunities to get creator bonuses (which does also bother me a bit).

Human engagement also has a way of snowballing. Presumably if instead of 80 bets by one bot the market had 20 bets by ten humans it would have gotten a lot more views in "the algorithm".

I appreciate that there's a strong intellectual current amongst Manifold leadership to let this kind of thing run whatever course is natural, but my sense is that if you polled the usership most people would say that it isn't good for the site.

@Yuna related question: do you think this restriction should apply to other markets with publicly-available/scrapeable probabilistic data? Or only sports?

@dglid Yes. Your implicit point - that Yuna’s logic actually applies to all (profitable) bot trading - is correct. The only difference is that sports produces by far the highest volume of events that would be fun to predict, and it’s a very short resolution horizon relative to other “topics”. So bots taking all the fake money in sports leads to a greater number of people going 😲

@dglid I don't think there should be a restriction. I think creators should have the option to disallow bot taking and thus incentivize them to post non-matching limit orders. Then I'd see what works best.

evanbd has suggested on discord to reduce subsidies if bot taking were disallowed. Interesting idea to me. https://discord.com/channels/915138780216823849/1194887929676443678/1195036347539411045

I'll just put in my vote for keeping it as-is; my hunch (have not analyzed the data) is that the rise of sports markets for bot capture have also led to a rise in other markets around sports that bots can't dominate, increasing the overall engagement with Manifold.

I agree with @BDStraw's comment about the purpose of Manifold being to surface accurate predictions/truth and the ways in that it can do that better than oddsmakers, even if only on the edges. If it's truly a significant barrier to engagement or user satisfaction with the product, then maybe Manifold should step in, but I'm not guessing there's evidence for that at this time.

I for one enjoy the way the Manifold UI shows the changes over time of sports probabilities (obviously I'm biased being so familiar with Manifold) and also have enjoyed trying to beat the bots from time to time (I've opened the betting UI to have my finger on the bet button, I've analyzed different video, streaming, and data feeds to see which ones are fastest, etc). I imagine it's frustrating for new users who might not realize at first that they're competing with bots who will usually have faster trigger fingers.

@dglid I think your first paragraph points to a lot of potential that sports can give to Manifold. But my strong guess is that that potential could be better realized without bot taking in sports. So for me, ideally creators have the option and then we can run experiments to see if engagement is better with or without bot taking.

@Yuna Yep all for experimentation and letting the data speak for itself, I'd just want to continue to move in the long-term direction of the team's self-stated goal (might have changed) of:

"Understanding what the future holds is a fundamental human need that influences all aspects of our behavior from how we choose to invest, to where we choose to work, to whom we choose to be around."

"We imagine a world where prediction markets are a ubiquitous feature of the public sphere—where every news article, boardroom meeting, or politician’s speech references a market forecast—and where the world is smarter as a result. .... the forecasts generated will be used by companies and individuals alike to make better decisions."

This goal is not achieved without accuracy. Maybe we can get minimally viable accuracy for these purposes without bots trading, especially with enough growth in the user base, but generally I wouldn't cut them off yet.

Really interested to see if I'll change my mind on this:

Please allow market creators to disallow bots on their markets. This is ridiculous! Nothing against you as a person @Yuna , I want to give you a high-five and a hug and go out for some tea. But @Yuna the bot? FUCK YOU.

@UniversalFC Go on discord, try to convince the team, if you're not doing that already. Maybe add some arguments to the profanities. 😅
You, me, and my engineer want the same.

@UniversalFC Sorry to myself for hijacking this thread, but you blocked me, so I can't message you on your own market where you said you accidentally mis-resolved something. It was in my favor, so cheers for that 😉 . But you can ask @SirSalty and I think mods to fix that for you. https://manifold.markets/UniversalFC/africa-cup-of-nations-group-stage-m#SQcFC4gr2fyGaOPzd8bL

Welcome to the leaderboard lol

@Joshua Thanks! 🙏 😊

is Yuna open source?

@machine No

"Of course, it depends on the purpose of the markets. If the purpose is accurate predictions, bots are an improvement. But the world doesn't need more accurate sports predictions. They are already publicly visible at every bookmaker. But Manifold needs engaged users."

Suppressing the best and fastest information in the name of user engagement would seem to me to be a pretty bad precedent for a number of reasons, definitely against that. I think Yuna teaches people who like these sports markets an important lesson - the publicly available real time info on sports prediction (bookies) is very, very good. As someone who bets on sports daily, I still enjoy trading on sports markets while games are in progress. Since manifold can't do all the things bookies do to hoard profits (freeze action during inflection points, charge vig), it gives you a perspective and opportunity you can't get just from staring at bovada.

@BDStraw I agree that epistemic humility is a good lesson to be learned from trying to beat bookie odds. But that lesson would eventually be learned without bot taking as well, because the successful humans are doing the same, just more manually. See @NamesAreHard previous activity and success on sports markets.

Idealistically, I follow your main point. But pragmatically I see huge potential for people to be drawn to Manifold by sports, but then get involved in more interesting/valuable corners of the question universe.

@Yuna mmm I guess I see your point and creators below who say there's a lot less engagement on their sports markets would seem to reinforce it. I would still say choosing a single "sector" to intentionally make the markets less efficient for user engagement purposes feels like a bad idea. But the evidence is already there that Yuna is bad for Manifold I guess. That's a shame.

@BDStraw No, Yuna is good for Manifold, because it will make Manifold give creators the option to disallow bot taking. 😉

@Yuna hahahaha certainly good for proving a point, the trick is in figuring out what the point is!

This is in a longterm market about Tour de France and I think it might be different for short term markets like NBA or NHL games but here’s a counterpoint to this proposal by @Lion and @Eliza:

@Yuna And as a creator, I'd like to be able to turn your taker orders OFF if my participants didn't want you there 🤣

@Eliza I want you to have the option to turn off bot taking 😊

I'm not sure this is really a problem.. Is the issue that people want to be able to bet faster and have more favourable odds? They should still be able to get "fair" odds, right?

The only issue I can think of atm is that someone puts a limit order at 50% and then a bot fills the whole thing just before the human sees their player fumble the ball or something. A human could probably do this too though

Please, tell me what I'm missing!

@Gen First, I've softened up a bit on this. But the main point I see is that if bots were forced to make the market, people could make much bigger bets. At a normal rip-off bookmaker, you can only do slightly negative expected value (-EV) bets, but at size! Let's say the true prob is 50%. They'll allow you to bet 1000s on YES at 45% (+EV for them, but you win now and then). Here, if you want to bet 1000s you drive the prob down to 0% without extra liquidity. That makes that bet VERY -EV. Then one of the bots scoops up that EV within seconds. If no bots were allowed at all, that misplaced bet would sit there as an easter egg for some human to find. If they find one, they will come back. If they find some regularly, people will see it in the profits and copy. 

Of course, it depends on the purpose of the markets. If the purpose is accurate predictions, bots are an improvement. But the world doesn't need more accurate sports predictions. They are already publicly visible at every bookmaker. But Manifold needs engaged users. Bots scooping up the easter eggs is bad for user engagement, I'm pretty confident about that. That includes @acc although that might be justified as a mana-sink. Still not sure if the trade-off is worth it. It's like Manifold builds this amazing maze with hidden easter eggs (mispriced questions) for people to scout for, and then builds a bot to collect the easter eggs themselves. @SG @JamesGrugett

Also, people hate losing against bots even more than against humans, I think: (They changed their username to "Antibot" after this. 😂 )

https://manifold.markets/FranklinBaldo/will-rayo-vallecano-beat-celta-vigo

Additionally, some market creators don't seem happy. See @NFL and @firstuserhere comments below and @KevinBurke . 

(Something that could make them happy again, might be to give them a cut of API fees generated on their markets though! Not sure if they are actually collected atm.)

On the other hand, some people welcome the increase in action that's possible with bots (@SemioticRivalry)

https://discord.com/channels/915138780216823849/1183489854206656593

(Pro-tip as a human who wants bigger bets at not too terrible prices: use limit orders. Eventually some bot will eat it up.)

I still think creators should have the .disallowBotTaking flag. Then they could experiment and see what works best. 

My prediction for long term user engagement:

BotsCanTake < NoBots << OnlyBotMaking

@Gen I’ve seen a decline on human activity on my NBA markets for example which makes them unprofitable to keep running. Sometimes the only two bettors are Ithaca and Yuna and I do not get any points for bot bets

@KevinBurke 💯

@Yuna Hmm.. my revised understanding

For traders

• bots provide liquidity, but only if you spam smaller bets that are big enough to trigger the bot to correct it, which is not really good liquidity

• Bots remove all possibility for a human to get an edge

Market makers

• bots don’t give trader bonuses

• Bots suck the fun out of the market so less people trade

Seem to cover it?

@Gen I would importantly distinguish between bot making and bot taking though. @ithaca seems to be doing a bit of making which makes the feed look terrible. (Because it’s harder to check the feed to see who bet what.)

But making (limit orders that don’t get matched immediately) adds value to everyone as far as I can tell. But it needs to be incentivized much more, it’s a lot harder to code because you have to manage positions and cancel bets etc.

So please don’t blame bots in general. But bot taking in particular. Baby, bathwater etc. 😊

(Feature request: toggle to hide non-matching trades or bot trades)

@Yuna This is a random not well thought out idea that likely nobody on the team approves of, BUT

what if we forced bots on sports markets to place minimum M100 limit orders that can’t be cancelled for, say 3-5 minutes? This means they have to lowball the odds a little but also provide liquidity - like a normal sports betting site provides

@Gen I’m not sure if you meant that instead of or in combination with banning taking for bots but I appreciate your brainstorming.

However, any time requirement over a second would render market making unprofitable in live events because a +EV yes bet at 99% can turn into a complete loss momentarily. (Score just before the final whistle for example).

Simply disallow taking for bots. Then they can figure out how to make making profitable.

@Yuna Yeah, that makes sense!

@Yuna FWIW I made a discord post suggestion: https://discord.com/channels/915138780216823849/1184704420391428186

@Gen Great, thanks!