greedbot.com
@greedbot-dot-com.bsky.social
Stop trading your life for a paycheck. Tested, systematic, Kelly-sized market research. The pros' math, in the open. Not financial advice.
https://www.greedbot.com
https://www.greedbot.com
Trading bot builders: one good training run proves little. This paper trained Bitcoin execution bots 100 times each. 11 of 100 plain runs learned to wait until forced to sell. Which fix looked best flipped between 30 and 100 runs.
arxiv.org/abs/2610.03369
arxiv.org/abs/2610.03369
October 6, 2026 at 8:22 PM
Trading bot builders: one good training run proves little. This paper trained Bitcoin execution bots 100 times each. 11 of 100 plain runs learned to wait until forced to sell. Which fix looked best flipped between 30 and 100 runs.
arxiv.org/abs/2610.03369
arxiv.org/abs/2610.03369
Which stock research do you read all the way through: newsletters, social threads, Reddit posts, or company filings?
October 6, 2026 at 4:45 PM
Which stock research do you read all the way through: newsletters, social threads, Reddit posts, or company filings?
Did an AI trader really pick winners? This paper tested one across 723 earnings reports. It made about 0.15% per event. Random picks with the same number of each kind of trade made about 0.26%. The gain came from stocks drifting up.
arxiv.org/abs/2610.04040
arxiv.org/abs/2610.04040
October 6, 2026 at 4:22 PM
Did an AI trader really pick winners? This paper tested one across 723 earnings reports. It made about 0.15% per event. Random picks with the same number of each kind of trade made about 0.26%. The gain came from stocks drifting up.
arxiv.org/abs/2610.04040
arxiv.org/abs/2610.04040
Mom's in the hospital with covid pneumonia. Could use your thoughts and prayers.
October 6, 2026 at 2:05 PM
Mom's in the hospital with covid pneumonia. Could use your thoughts and prayers.
Nasdaq 100 pizza going into Tuesday: Micron takes the top slice at 2.60%, up 0.29pp. Astera Labs and Microsoft gained next. Exelon and Ross Stores lost the most.
Model target weights are not expected returns. Research only. Start free ranking.
Model target weights are not expected returns. Research only. Start free ranking.
October 6, 2026 at 12:57 PM
Nasdaq 100 pizza going into Tuesday: Micron takes the top slice at 2.60%, up 0.29pp. Astera Labs and Microsoft gained next. Exelon and Ross Stores lost the most.
Model target weights are not expected returns. Research only. Start free ranking.
Model target weights are not expected returns. Research only. Start free ranking.
AI-written trading bots can pass a backtest and still drift from your risk rules. This MintEval paper compared their trades bar by bar with the intended strategy. The best model tested drifted on about 1 in 4 tasks.
arxiv.org/abs/2610.03080
arxiv.org/abs/2610.03080
October 5, 2026 at 8:22 PM
AI-written trading bots can pass a backtest and still drift from your risk rules. This MintEval paper compared their trades bar by bar with the intended strategy. The best model tested drifted on about 1 in 4 tasks.
arxiv.org/abs/2610.03080
arxiv.org/abs/2610.03080
What did your trading bot or script do recently that made you laugh or swear?
October 5, 2026 at 4:45 PM
What did your trading bot or script do recently that made you laugh or swear?
Order book research: a tiny model that sees every order placed, canceled, and filled scored best in 11 of 12 short-term price direction tests and ran 5 to 8 times faster than far bigger models.
arxiv.org/abs/2610.02917
arxiv.org/abs/2610.02917
October 5, 2026 at 4:22 PM
Order book research: a tiny model that sees every order placed, canceled, and filled scored best in 11 of 12 short-term price direction tests and ran 5 to 8 times faster than far bigger models.
arxiv.org/abs/2610.02917
arxiv.org/abs/2610.02917
Dow pizza going into Monday: Nvidia still leads at 6.51%. Nike and Boeing gained the most weight. J&J and Microsoft lost the most. Surprised by any of those?
Model target weights are not expected returns. Research only. Start free ranking.
Model target weights are not expected returns. Research only. Start free ranking.
October 5, 2026 at 12:57 PM
Dow pizza going into Monday: Nvidia still leads at 6.51%. Nike and Boeing gained the most weight. J&J and Microsoft lost the most. Surprised by any of those?
Model target weights are not expected returns. Research only. Start free ranking.
Model target weights are not expected returns. Research only. Start free ranking.
Profit-only RL bots ride up-drifts. Hard loss penalties freeze them in chop.
This PPO paper adds a regime hint to keep upside and still limit drawdowns.
arxiv.org/abs/2610.01325
This PPO paper adds a regime hint to keep upside and still limit drawdowns.
arxiv.org/abs/2610.01325
October 4, 2026 at 8:22 PM
Profit-only RL bots ride up-drifts. Hard loss penalties freeze them in chop.
This PPO paper adds a regime hint to keep upside and still limit drawdowns.
arxiv.org/abs/2610.01325
This PPO paper adds a regime hint to keep upside and still limit drawdowns.
arxiv.org/abs/2610.01325
What do you use your phone for in your trading or investing routine?
October 4, 2026 at 4:45 PM
What do you use your phone for in your trading or investing routine?
Traders bid what they would take for holdings they already have. Repeat that, and auction prices can settle.
A paper on double auctions maps that loop to a stable clear.
arxiv.org/abs/2610.01562
A paper on double auctions maps that loop to a stable clear.
arxiv.org/abs/2610.01562
October 4, 2026 at 4:22 PM
Traders bid what they would take for holdings they already have. Repeat that, and auction prices can settle.
A paper on double auctions maps that loop to a stable clear.
arxiv.org/abs/2610.01562
A paper on double auctions maps that loop to a stable clear.
arxiv.org/abs/2610.01562
Sunday crypto pizza movers: JUST dropped overnight. Quant still leads. PAX Gold climbed to third.
Quant 4.14%. JUST 3.00% (down 0.25pp). Bitcoin 2.85%.
Model target weights are not expected returns. Research only. Start free ranking.
Quant 4.14%. JUST 3.00% (down 0.25pp). Bitcoin 2.85%.
Model target weights are not expected returns. Research only. Start free ranking.
October 4, 2026 at 12:57 PM
Sunday crypto pizza movers: JUST dropped overnight. Quant still leads. PAX Gold climbed to third.
Quant 4.14%. JUST 3.00% (down 0.25pp). Bitcoin 2.85%.
Model target weights are not expected returns. Research only. Start free ranking.
Quant 4.14%. JUST 3.00% (down 0.25pp). Bitcoin 2.85%.
Model target weights are not expected returns. Research only. Start free ranking.
Passive crypto liquidity can trail the pool average. Active liquidity moves around trades. A paper found a wider gap on Ethereum than on faster chains.
arxiv.org/abs/2609.37963
arxiv.org/abs/2609.37963
October 3, 2026 at 8:22 PM
Passive crypto liquidity can trail the pool average. Active liquidity moves around trades. A paper found a wider gap on Ethereum than on faster chains.
arxiv.org/abs/2609.37963
arxiv.org/abs/2609.37963
Traders with similar models can send different orders off one headline. One paper saw a 48 point swing in a model share of orders. Net count barely moved.
arxiv.org/abs/2610.01897
arxiv.org/abs/2610.01897
October 3, 2026 at 6:02 PM
Traders with similar models can send different orders off one headline. One paper saw a 48 point swing in a model share of orders. Net count barely moved.
arxiv.org/abs/2610.01897
arxiv.org/abs/2610.01897
Saturday crypto ranking: Quant leads. Bitcoin is fifth.
Quant 4.14% target weight. Bitcoin 2.85%.
Model target weights are not expected returns. Research only. Start free ranking.
Quant 4.14% target weight. Bitcoin 2.85%.
Model target weights are not expected returns. Research only. Start free ranking.
October 3, 2026 at 5:56 PM
Saturday crypto ranking: Quant leads. Bitcoin is fifth.
Quant 4.14% target weight. Bitcoin 2.85%.
Model target weights are not expected returns. Research only. Start free ranking.
Quant 4.14% target weight. Bitcoin 2.85%.
Model target weights are not expected returns. Research only. Start free ranking.
Bot builders: QuantCode trains language models to generate executable algo-trading code traders can run and check.
arxiv.org/abs/2609.39420
arxiv.org/abs/2609.39420
October 2, 2026 at 8:22 PM
Bot builders: QuantCode trains language models to generate executable algo-trading code traders can run and check.
arxiv.org/abs/2609.39420
arxiv.org/abs/2609.39420
Morning before charts: coffee, news, workout, or jump in?
October 2, 2026 at 4:45 PM
Morning before charts: coffee, news, workout, or jump in?
Big buy, smaller price shove than a linear model expects: hidden liquidity and flatter impact as size grows.
arxiv.org/abs/2609.37872
arxiv.org/abs/2609.37872
October 2, 2026 at 4:22 PM
Big buy, smaller price shove than a linear model expects: hidden liquidity and flatter impact as size grows.
arxiv.org/abs/2609.37872
arxiv.org/abs/2609.37872
BAC or WFC this week? Optimistic (+71%) vs Pessimistic (-31%) on the weekly model.
Model percentages are not expected returns. Research only. Start free ranking.
Model percentages are not expected returns. Research only. Start free ranking.
October 2, 2026 at 12:57 PM
BAC or WFC this week? Optimistic (+71%) vs Pessimistic (-31%) on the weekly model.
Model percentages are not expected returns. Research only. Start free ranking.
Model percentages are not expected returns. Research only. Start free ranking.
Grading a trading bot stock research? FinAutoRubric turns expert guidance into grading rules for each question.
Preprint: arxiv.org/abs/2609.35744
Preprint: arxiv.org/abs/2609.35744
October 1, 2026 at 8:22 PM
Grading a trading bot stock research? FinAutoRubric turns expert guidance into grading rules for each question.
Preprint: arxiv.org/abs/2609.35744
Preprint: arxiv.org/abs/2609.35744
Who pulls you away from your charts?
October 1, 2026 at 4:45 PM
Who pulls you away from your charts?
Reading an institution's stock commentary? Check its positions too. Say / Echo / Do compares statements, media coverage and positions in a simulated market.
Preprint: arxiv.org/abs/2609.38545
Preprint: arxiv.org/abs/2609.38545
October 1, 2026 at 4:22 PM
Reading an institution's stock commentary? Check its positions too. Say / Echo / Do compares statements, media coverage and positions in a simulated market.
Preprint: arxiv.org/abs/2609.38545
Preprint: arxiv.org/abs/2609.38545
Synopsys (SNPS): GreedBot's weekly view moved from Pessimistic (-100%) to Optimistic (+42%).
Model percentages are not expected returns. Research only. Start free ranking.
Model percentages are not expected returns. Research only. Start free ranking.
October 1, 2026 at 12:57 PM
Synopsys (SNPS): GreedBot's weekly view moved from Pessimistic (-100%) to Optimistic (+42%).
Model percentages are not expected returns. Research only. Start free ranking.
Model percentages are not expected returns. Research only. Start free ranking.
Most LLM trading tests only check whether a bot guesses stock direction. This preprint puts agents into option markets with nonlinear payoffs and multi-leg structures - useful if you are building bots that must size hedges, not just call up or down.
Preprint: arxiv.org/abs/2609.33470
Preprint: arxiv.org/abs/2609.33470
September 30, 2026 at 8:22 PM
Most LLM trading tests only check whether a bot guesses stock direction. This preprint puts agents into option markets with nonlinear payoffs and multi-leg structures - useful if you are building bots that must size hedges, not just call up or down.
Preprint: arxiv.org/abs/2609.33470
Preprint: arxiv.org/abs/2609.33470