@GeorgeVii Sure! I haven’t done anything clever, and there’s just a bunch of disorganized Jupyter notebooks. But the most helpful thing for me has been to have a spreadsheet of my own subjective probabilities and then have a script compare them to the market and output the ones with the highest expected return. I’ve been using something similar for the main Manifold with returns per day too. But since the Salem markets are mostly long-term, I’ve ignored the length of markets. Here’s the Python code:

import requests

import csv

BASE_URL = "https://salemcenter.manifold.markets/api/v0"

# Read subjective guesses from csv file

# {'table_name': 'probability', ...}

table = {}

with open('myprobs.csv', newline='') as csvfile:

    reader = csv.DictReader(csvfile)

    for row in reader:

        table[row["slug"]] = float(row["probability"])

# Get a list of all markets

req = requests.get(BASE_URL + "/markets")

markets = req.json()

# Join markets with csv table

for market in markets:

    market['slug'] = market['url'].split('/')[-1]

    if market['slug'] in table:

        market['my_prob'] = table[market['slug']]

    else:

        print(f'No probability for {market["slug"]}')

markets = [m for m in markets if 'my_prob' in m]

def get_returns(their_prob, my_prob):

    # WLOG my_prob <= their_prob

    if their_prob > my_prob:

        their_prob = 1 - their_prob

        my_prob = 1 - my_prob

    return my_prob / their_prob

for market in markets:

    market['returns'] = get_returns(market['probability'], market['my_prob'])

markets.sort(key=lambda x: x['returns'], reverse=True)

for market in markets[:10]:

    nice_string = f"Return {market['returns']-1:.1%}, {market['probability']:.1%} => {market['my_prob']:.1%}: {market['question']}"

print(nice_string)

And here are my current subjective probabilities (myprobs.csv). They could probably use an update:

slug,probability,source

supreme-court-ban-race-in-college-a,0.75,"https://www.metaculus.com/questions/9573/scotus-ends-racial-preferences-in-admissions/"

will-elon-musk-buy-twitter,0.55,"https://www.metaculus.com/questions/10662/will-elon-musk-buy-twitter-in-2022/"

republicans-favored-by-summer-2023,0.7,"Base rate is the other side wins midterms"

chinese-military-action-against-tai,0.06,"Skirmsihes happen (e.g. https://en.wikipedia.org/wiki/2020%E2%80%932022_China%E2%80%93India_skirmishes), but nothing in a long while (https://en.wikipedia.org/wiki/Template:PRC_conflicts). Also https://www.metaculus.com/questions/7227/armed-conflict-chinataiwan-by-2024/ suggests <10%"

will-trump-announce-in-2022,0.48,"Trump announced originally on16 June 2015, so this would be earlier than last time. Biden April 25, 2019"

will-republicans-win-michigan-3rd-d,0.42,"https://projects.fivethirtyeight.com/2022-election-forecast/house/michigan/3/"

biden-cancel-student-debt-in-2022,0.46,"https://kalshi.com/events/SDEBT/markets/SDEBT-23JAN01, though different resolution criteria"

mask-mandate-in-any-state-by-end-of,0.32,"https://www.metaculus.com/questions/12001/us-state-mask-mandate-before-212023/, but has different date"

covid-falls-below-20000-cases-in-20,0.2,"Things go up in winter, but testing could widely stop or something weird happen. Lower bound 24% from Kalshi https://kalshi.com/events/CASE7DA-22/markets/CASE7DA-22-A11.7"

over-100000-monkeypox-cases-in-2022,0.6,"Things still going strong in the US but increase in other countries. No great existing forecast, but assuming 20% in the US https://www.metaculus.com/questions/10976/total-estimated-monkeypox-infections-in-2022/ give about 63%"

russia-annex-territory-by-july-2023,0.73,"Just defaulting to market"

russia-annex-territory-in-2022,0.54,"Just defaulting to market"

world-gdp-growth-3-or-more-in-2022,0.44,"https://kalshi.com/events/GDPW-22/markets/GDPW-22-A3"

china-gdp-growth-4-or-more-in-2022,0.3,"https://kalshi.com/events/GDPCN-22/markets/GDPCN-22-A4"

will-russia-control-kramatorsk-on-o,0.5,""

will-russia-control-kramatorsk-on-5,0.55,""

covid-peak-of-over-300000-by-end-of,0.3,"Peak last time was 800,000 but after new years. US now only has 600,000 tests performed a day, but that fluctuates"

us-gdp-growth-over-4-in-any-quarter,0.12,"For Q3 Kalsi doesn't even consider >4% (us-gdp-growth-over-4-in-any-quarter), for Q4 no numbers yet. But this would be a big change."

russiaukraine-ceasefire-by-7312023,0.5,"Metaculus has 14% until end of year (https://www.metaculus.com/questions/10050/bilateral-ceasefire-between-russia--ukraine/), but this is twice the time. Plus there is precedent (Minsk, corn export)"

will-russia-control-kherson-on-5312,0.6,"Copied from market"

will-russia-control-kherson-on-1031-73b8bcce2a70,0.65,"Copied from market"

will-republicans-win-the-senate-in-d755992df1b5,0.35,"https://www.metaculus.com/questions/11422/oz-to-win-2022-pa-senate-race/"

will-republicans-win-the-senate-in,0.4,"https://metaforecast.org/?query=senate+arizona"

us-gdp-growth-1-or-more-in-2022-q3,0.62,"https://kalshi.com/events/GDP-22SEP30/markets/GDP-22SEP30-T1.0"

biden-the-favorite-in-summer-2023,0.7,"Mostly, will he be running? Incumbents are pretty much automatically the frontrunners. Also, the usual primaries won't have started yet at this point."

will-republicans-win-the-senate,0.45,"https://www.metaculus.com/questions/5632/gop-controls-us-senate-in-2023/, https://projects.fivethirtyeight.com/2022-election-forecast/senate/"

will-republicans-win-the-house-of-r,0.8,"https://www.metaculus.com/questions/5814/2022-us-house-midterm-election/"

will-donald-trump-be-indicted-for-a,0.4,"This is currently at 36% on Polymarket https://polymarket.com/market/will-donald-j-trump-be-indicted-by-december-31-2022 and about 32% on Smarkets https://smarkets.com/event/42769683/politics/us/donald-trump/2022/09/02/00-00/donald-trump-to-be-indicted-in-2022"

will-donald-trump-be-indicted-for-a-411b63e42092,0.6,"Extrapolate the above"

china-covid-wave-by-winter,0.48,"Uncertain but Metaculus says low https://www.metaculus.com/questions/10936/china-to-reach-250k-covid-cases-per-day/"

will-republicans-win-the-senate-in-836641ae8816,0.5,"https://metaforecast.org/?query=senate+nevada"

biden-at-40-approval-on-labor-day-w,0.62,"https://projects.fivethirtyeight.com/biden-approval-rating/. Seems like 40% is about the current value so probably 50%, but deferring to the market somewhat"

republicans-favored-in-senate-on-la,0.25,"https://projects.fivethirtyeight.com/2022-election-forecast/senate/"

will-republicans-win-the-senate-in-5dd0dba203a5,0.73,”https://metaforecast.org/?query=senate+ohio"

@BionicD0LPH1N Decided to tip M$50 instead to signal I'll tip even ideas that I don't necessarily agree with.

@GeorgeVii As in, after you make a trade I make the same trade in the same direction? I'm not suspicious at all that you'd try to game this... absolutely not :P

@TheBayesian That ended up being super helpful, so if you comment somewhere I'll tip you M$80 and potentially more based on how well it ends up working. Tipping @1941159478 for similar reasons.

@1941159478 Gib code 🙏

ʕ•́ᴥ•̀ʔっ

@GeorgeVii tipping because cute

@BionicD0LPH1N Aww I'm flattered!! <3

@1941159478 LOL three ML models in a trench coat is a hilarious way to phrase things.

Anyway, I had simply noticed George Vii doing something similar (increasing layered limit orders I mean) so I thought it would be a great idea to try it out, because I know he’s really good. I appreciate the compliment greatly, but you should know that he should get all the credit. 😊

@BionicD0LPH1N Ah, cool, thank you so much for following up! It’s good we know for sure now. Happy coding to you then

I actually already thought you were three ML models in a trench coat because some of your limit orders are so beautifully staggered 😊

@BionicD0LPH1N Update: apparently the way you’ve been using it is allowed, so you’re right on everything.

@1941159478 Ahh I was wondering whether you were using it because you were so quick to bet against beginners! I've only been doing it through completely unhealthy practices :P

James said on discord:
"Actually I take it back, bots are probably not allowed (or use of the API) because it gives an advantage to programmers, and this is intended as a general forecasting contest."

I'm a bit annoyed that they haven't announced it anywhere except to me when I explicitly asked for it. It creates the weird situation where we're strongly disincentivized to asked about what is allowed. In fact, I'm super annoyed at this tournament for obviously not being a forecasting tournament. I don't know what proportion of the profits you made were through being good enough to distinguish 50% chance to 45% chance, but for me it's basically none of it. Rant over.

I don't think you should report yourself somewhere nor that you should be penalized for it, it's literally not said anywhere except on a single channel in the discord. I'll talk to James about maybe making an announcement about it, or something like that.

@BionicD0LPH1N Only parts of the API are disabled: https://github.com/manifoldmarkets/manifold/compare/salemcenter

I didn’t see anything about the remaining API being forbidden, only to not “hack” the tournament. If it is, I’ll definitely have to report myself somewhere 😬

@TheBayesian They say using the API is not allowed, so I don't know how I would do that without cheating? There's the standard notifs from manifold that I can activate to my email, but that sounds like a pain

@BTE That's a pretty good point, although I can't really test it fully (I already have money spread in multiple markets and I'm not willing to sell rn). But I'm going to update my strategy slightly to trade if fewer markets, and I'll see how it goes. Tipping S$15 for now, potentially more later.

@BionicD0LPH1N Currently, I have 350 in one position and even at that scale the value of the position has barely (less than 3 percent) changed after a 22 percent move in the market. If you spread money thin you will need like 20 percent movement to make 1 percent gain. Maybe I am not paying close enough attention, but I have been shocked by how little the value of my bets is changing with changes in the market probability.

@BTE That seems like a bad idea to me. Trading in many markets is good because it's less risky and you move the market price less which reduces fees. What's the argument in favor of doing this?

Give me money!!! LOL!! 🤑🤑🤑

Reminder to comment somewhere if you want to be tipped!