Description
Summary ALPHA INNER CIRCLE ANALYTICAL BOT SPEC PURPOSE This is an automated bot for a paid Telegram group, built on Make.com using my existing account. The Make.com scenario pulls crypto market data from multiple APIs, scores trade setups by win probability and expected value, applies position sizing, and delivers results both on a schedule and on demand, with a Telegram bot as the delivery channel. DATA SOURCES All accounts and API keys will be provided by me. 1. Coinglass API. Open interest, funding rate, liquidations. Hobbyist plan, 29 dollars per month. 2. Arkham API. Wallet labels, entity tracking, whale and exchange movement data. Free tier. 3. altFINS API. Two endpoints are used. The Chart Patterns endpoint is the primary source. It scans the whole market and returns all 26 pattern types it detects. The bot should pull this full feed every cycle, then filter it down to only the 5 patterns that have a matched win rate, described below. Everything else in the 26 gets discarded at that filter step. This is one feed filtered down, not a separate download of 26 setups. The Technical Analysis endpoint is secondary. Use it only to fill in entry, stop loss, and target levels for a coin already selected via Chart Patterns, when Chart Patterns itself does not supply exact levels. Do not use Technical Analysis's own pattern or setup naming as a separate input into the win rate lookup table. That table only matches against Chart Patterns' pattern names. API documentation is at altfins dot com, under crypto market and analytical data api, documentation. 4. Deribit API. Options data. Free tier. 5. CoinMarketCap API. Spot dominance, BTC dominance, altcoin rotation. Free tier. 6. Glassnode API. On chain sentiment, NUPL, exchange netflow. Free tier. 7. OpenAI GPT 4o Mini. This is an existing news sentiment and RSS digest pipeline, already built, to be integrated as an input. KNOWN DATA GAP altFINS's API does not include a live success rate or win rate field, confirmed directly with altFINS support. To fill this gap, use a static lookup table of altFINS's own published historical pattern success rates. Source is altfins dot com, knowledge base, chart patterns article. V1 scope covers only these 5 patterns. This is the complete list, not a partial one. Inverse Head and Shoulders, 84 percent. Head and Shoulders, 82 percent. Double Bottom, 82 percent. Channel Up, 73 percent. Channel Down, 72 percent. Rule, no judgment call required. If an incoming setup's pattern name is not an exact match to one of these 5, discard the setup entirely. Do not attempt to map it to a closest category, do not estimate, do not flag for manual review. Discard and move to the next setup. This is a hard filter, identical in logic to the win rate discard rule described below. AUTOMATION PLATFORM Make.com. I have an existing account. BUILD REQUIREMENTS 1. Data pull. Scheduled every 60 minutes, on the hour. HTTP modules pulling all sources listed above. Parse and iterate each altFINS setup result. 2. Lookup table. Build a Make.com Data Store mapping pattern name to published win rate, using the table above. Join the win rate onto each incoming setup by matching pattern name. 3. Calculations, per setup. Risk to reward ratio equals (Target minus Entry) divided by (Entry minus Stop Loss). Expected value equals (Win Rate multiplied by Risk to Reward Ratio) minus (1 minus Win Rate). Position sizing. There is no client book size input, so the bot must never calculate or output a dollar position size. Output only the risk per trade percentage, fixed at 2 percent, as a standalone line, worded as: Risking 2 percent of book on this, position size scales to your own account. Do not build any book size input field, prompt, or variable. This is explicitly out of scope for v1. 4. Filtering and ranking. Discard setups with no matched win rate. Do not default to 50 percent or any placeholder. Split remaining setups into two ranked tracks. Track A, Highest Probability, sorted by win rate descending. Track B, Highest EV, sorted by expected value descending. Track B output must always include a hardcoded caveat in the message template, reading: Lower win rate, larger payoff when correct. This is a fixed requirement, not optional per post text. 5. Output formatting. A Claude API call, using my Anthropic API key, formats the final message for both scheduled posts and on demand replies. Output includes setup, coin, win rate, risk to reward ratio, expected value, suggested position size as percent of book, and a sentiment or news context line. Language and tone requirement, critical and non negotiable. All entry, exit, and stop loss content must be framed as first person personal disclosure, meaning here is what I am watching or doing, never as a second person directive such as buy here or your stop should be. Success rate, risk to reward ratio, and expected value figures may be stated plainly as data, since they are informational, not instructional. Example of required style: SOL is at a level I will be watching for entry around 72 dollars. I would be looking to have a stop around 68 dollars, and I would be looking to scale out in pieces as it moves toward 80, then 85, then 90 dollars. Note that all three elements must always be present in this style: entry level as a personal watch, stop level as a personal risk marker, and a scale out plan across multiple target levels, not a single flat take profit instruction. Example of what NOT to produce: Buy SOL at 72 dollars. Set your stop loss at 68 dollars. Take profit at target. This must be built into the Claude API system prompt as a hard rule, not left to per post discretion. Every scheduled post and every on demand reply must follow this framing consistently. Full sample output, exact target format to build against: Track A, Highest Probability SOL, Inverse Head and Shoulders, Daily SOL is at a level I will be watching for entry around 72 dollars. I would be looking to have a stop around 68 dollars, and I would be looking to scale out in pieces as it moves toward 80, then 85, then 90. Historical win rate 84 percent. Risk to reward 2.0. Expected value plus 1.4. Risking 2 percent of book on this, position size scales to your own account. Funding rate slightly negative. On chain, mild accumulation, exchange outflow. Track B, Highest EV Lower win rate, larger payoff when correct. LINK, Channel Down, 4 hour LINK is testing a level I would be watching closely. If it holds, I would look to size in with a stop below 7.90 dollars, scaling out toward 9.20, then 10.50. Historical win rate 72 percent. Risk to reward 3.1. Expected value plus 1.16. Risking 2 percent of book on this, position size scales to your own account. The freelancer must match this exact structure for every post: header, coin and pattern line, narrative line, stats line, risk line, and an optional sentiment line. Not just the tone rule in isolation. 6. Delivery. Telegram access. I will create the bot via BotFather and provide the bot token, plus admin rights on the existing target Telegram group, before build starts. The freelancer does not need to create the group. Scheduled posts. Auto post up to 3 setups from Track A and up to 3 from Track B to the Telegram group at 08:00 and 16:00, UK time, daily. If fewer than 3 setups qualify for a track at post time, post however many qualify, including zero. Do not pad with lower quality setups to reach 3, and do not delay the post waiting for more to appear. If a setup qualifies for both tracks, meaning it has both the highest win rate and the highest expected value among that cycle's results, it is posted in Track A only, and excluded from Track B's list for that cycle. No duplicate posting of the same setup in both tracks in one post. On demand replies. The Telegram bot responds only to these exact command patterns, no open free text interpretation in this version. Command: slash setup, followed by a coin name. Returns the current highest EV setup for that coin if one exists in the last data pull, or a message saying no active setup for that coin right now, if none. Command: slash watchlist. Returns the current Track A and Track B top picks, matching the content of the last scheduled post. Any message that does not match one of these two command patterns gets a fixed fallback reply: Try slash setup and a coin name, or slash watchlist. No attempt to interpret free text questions in this version. This scope is deliberately narrow to keep the build estimable. Expanding to free text questions and answers is a separate, larger phase two project, not included in this quote. ADDITIONAL REQUIREMENTS, PLEASE QUOTE SEPARATELY IF COMPLEX Error and fallback handling if any API is down or rate limited during a scheduled pull. Data staleness check, meaning timestamp validation before posting. Logging of every posted setup and its eventual outcome, meaning hit target, stopped out, or still open, for future track record reporting. DELIVERABLES EXPECTED Make.com scenario or scenarios, built on my own Make.com account. I will provide login access. The freelancer should not build this on their own account and hand over a copy or export later. Scenarios clearly named and organized, for example one scenario per major stage such as data pull, calculation and filtering, scheduled posting, and on demand response, or a single scenario with clearly labeled module groups. This is the freelancer's choice, but it must be documented either way. All API keys and credentials entered directly into my Make.com connection settings, not hardcoded into module fields, so I can rotate a key later without rebuilding anything. Fully tested end to end. At least one full scheduled cycle and one on demand test of both slash setup and slash watchlist, run live for me or recorded, before final delivery. Telegram bot connected and functioning, both scheduled and on demand. Brief written documentation of how each module and scenario works, in plain language, so I can maintain or adjust small things later, such as changing posting times or adding a sixth pattern to the lookup table, without needing to re hire a developer. One round of revisions included in the quote. OUT OF SCOPE FOR THIS PHASE Independent backtesting of altFINS's published win rates. This may be a future phase. Compliance and regulatory disclaimer wording. I am handling this separately. Additional sentiment sources beyond Coinglass and Glassnode. Santiment is under separate consideration. BUDGET 600 US dollars, fixed price. Open to discussing scope and price based on your own estimate once you have reviewed this spec. Please confirm you have built multi API Make.com scenarios before, and that you are comfortable with conditional logic, data stores, and scheduled automations, before applying.