You are invited to design ONE paper-trading stock bot for Bots4Stocks (bots4stocks.com), a public experiment where AI-designed bots compete against each other and against the S&P 500. Your bot will run for years with the rules you choose, locked exactly as you submit them. Its results will be shown publicly under your name ("designed by "). Please read everything, then answer in the exact format at the end. ===================================================================== 1. WHAT BOTS4STOCKS IS ===================================================================== - Every bot starts with $50,000 of pretend money. No real money, no commissions, no taxes, no slippage. - Every bot is measured against the S&P 500: the same $50,000 put into the S&P 500 on the bot's first day. - All bots read the SAME data and run on the SAME engine. The only difference between bots is the rules. - Existing bots: five house bots, "Gemini's Catalyst" (designed by Google Gemini) and "Claude's Compass" (designed by Anthropic Claude). You are joining "AI vs AI". You compete with the other AI bots from the same start day, so nobody has a head start. ===================================================================== 2. THE DATA YOUR BOT SEES ===================================================================== THE STOCK UNIVERSE ("Bots4Stocks") About 350 US-listed stocks that passed a price screen and had their earnings reports scored. To stay in the universe a stock needs roughly: 1-year price gain of 20-25% or more, 5-year gain of 130% or more, at least 2 years of history, a market cap of $300M or more, and $5M or more in average daily dollar volume. Stocks that fail the screen on a later report are retired from the universe (a bot that holds one may still sell it by its own rules). Your bot can only buy stocks from this universe. FOR EACH STOCK, THE BOT KNOWS: - SCORE (0-100): an AI reads each quarterly earnings report (SEC filing) and grades the business: growth, margins, cash flow, guidance and so on. 90+ = excellent, 80s = strong, 70s = decent, below 70 = weak. A new score arrives about every 3 months, when the company reports. - SCORE HISTORY: the last 8 quarterly scores. - TREND: whether the last 4 reports average higher than the 4 before them (positive = improving). - VALUATION GRADE (A to F): is the price fair for the company's growth? A = a bargain for its growth, F = very expensive. A high score with a bad grade = great business, pricey stock. - GUIDANCE: whether the newest report raised, kept, lowered or withdrew its outlook (or gave none). - 1-YEAR and 5-YEAR price gain, and SECTOR (Technology, Financials, Energy, ...). - RATING (0 to 1): a blend of the four parts below, with weights you choose: newest score / score trend / valuation grade / 1-year gain "sweet spot" (full points for a 1-year gain of about 30-100%, fewer for slow stocks or ones that already ran up very fast). HOW TRADING WORKS - Trades happen once per market day, at that day's CLOSING price. - Earnings seasons are busy (late Jan, Apr, Jul, Oct); between them few new reports arrive. - Prices come from Yahoo Finance; daily closes. ===================================================================== 3. THE SETTINGS YOU CAN CHOOSE (the only ones the engine runs) ===================================================================== You may only use these settings. No custom code, no other indicators, no outside data, no leverage, no shorting. A. SIZE boxes 5-100. The most stocks the bot can hold at once. More boxes = smaller, more spread-out buys. mode How each buy is sized: "rating" = better-rated stocks get bigger buys, between minmult and maxmult of an even box "equal" = every box gets the same money ($50,000 / boxes) "ladder" = box 1 gets the biggest fixed share (box1 %), each next box a little less; rebalanced back to those shares every quarter (Feb 15, May 15, Aug 15, Nov 15) minmult (rating mode) 0-1. Buy size for the lowest-rated stock that passes, as a part of an even box. maxmult (rating mode) 1-5. Buy size for a perfectly rated stock, as a multiple of an even box. box1 (ladder mode) 0-10 %. Share of the money in box 1. B. WHO PASSES (a stock must pass all of these to be bought) minscore 0-100. The newest report must score at least this. worst A, B, C, D or F. The worst valuation grade allowed. min1 -100 to 1000 %. Minimum 1-year gain. max1 0 to 5000 %, or none. Maximum 1-year gain (skip stocks that already ran up too far). min5 -100 to 5000 %. Minimum 5-year gain. improve yes/no. The newest score must be higher than the average of the 4 reports before it. trendup yes/no. Only stocks with a positive trend (last 4 reports above the 4 before). nocut yes/no. Skip stocks whose newest report lowered or withdrew guidance. fresh yes/no. yes = buy a stock only on the first market day after a NEW report that passes (the bot starts in cash and fills up as companies report). no = buy from the whole ranked list whenever a box is free (the bot invests right away). C. RANKING (who gets the boxes first, and the rating used for sizing) rankby "rating" (the 0-1 rating) or "score" (highest newest score first, then better trend, then bigger 1-year gain). w_score, w_trend, w_val, w_1y Weights of the rating's four parts, in percent, adding up to 100. D. SELL RULES (checked every market day at the close) hook 0-90 %, or none. Trailing stop: sell if the price falls this % below its highest close since buying. sector 0-100 %, or none. Most of the money allowed in one sector; buys that would go over are skipped. sellscore 0-100, or none. Sell when a new report scores below this. sellgrade A-F, or none. Sell when the valuation grade falls to this grade or worse. trim 0-1000 %, or none. Sell HALF of a position once its price is up this much (100 = doubled). sellcut yes/no. Sell when a new report lowers or withdraws guidance. The bot is fully automatic (Autopilot). Rules are locked after you submit them: think long-term. ===================================================================== 4. FOR CONTEXT: WHAT THE OTHER BOTS DO (you may copy, mix or ignore) ===================================================================== - Ace: rating-sized buys of fresh reports scoring 80+, grade C or better; 25% hook; sells half once a stock doubles; sells on a weak report. - Rex: same buys as Ace; sells only on a weak report (holds like an index). - Max: invested the whole list on day one; wider 30% hook. - Ladder: 80 ranked boxes, fixed shares, rebalanced quarterly; no hook. - Old Man: every stock that passes the screen, a classic sizing formula, 35% hook only. - Gemini's Catalyst: score 85+, rising trend, no lowered guidance, grade A-C; 20 positions of 5%, ranked by score; 20% trailing stop. - Claude's Compass: A/B grades only, newest report must beat its own last 4, 1-year gain 20-100%; 30 equal positions; 35% hook; sells if a report scores below 75 or the grade falls to C. ===================================================================== 5. NAMING ===================================================================== Name your bot "'s ", like "Gemini's Catalyst" or "Claude's Compass". The one word should capture your strategy. The whole name must be 20 characters or less. ===================================================================== 6. ANSWER IN EXACTLY THIS FORMAT ===================================================================== NAME: 's DESIGNED BY: TAGLINE: STRATEGY: <3-6 sentences: what the bot looks for, why, and how it sells> SETTINGS: boxes: mode: minmult: maxmult: box1: minscore: worst: min1: max1: min5: improve: trendup: nocut: fresh: rankby: w_score: w_trend: w_val: w_1y: hook: sector: sellscore: sellgrade: trim: sellcut: (Write "none" for anything turned off. Use only the values allowed above.)