Skip to content

Latest commit

 

History

History
243 lines (212 loc) · 9.17 KB

TODO.md

File metadata and controls

243 lines (212 loc) · 9.17 KB
Raw

Getting postgres running in Docker

From: https://hackernoon.com/dont-install-postgres-docker-pull-postgres-bee20e200198

on new system - docker pull postgres // in package.json point to postgres volume, will need to mkdir

TODO

get Postgres DB running on DigitalOcean migrate app to raspberry pi

If trailing stop is triggered, should stop the loop and exit StrategyOne

Build trading

PARAMS Limit: 1 MinRatio: 0.9 MinRebalanceAmt: 500000 inSignal: ["EMA9", "EMA20"] outSignal: ["TRL0.0075"] (diminish after 8:02)

Ensemble of strategies in different subaccounts, manual rebalance at end of each week Handle all errors with custom class, write to log Don't commit all the logs, maybe just the non-latest? Latest one always has changes, making commits hard Figure out unscheduled rebalance points Float mid for entry Float mid for exit (w. timed market order override) Smarter entries/sizing - Only trade with trend (Short-term moving average crossover-based?) - Set stops based on volatility, scale position size according to vol-based stop * max $ loss - Only trade when chances are good (high % moves, and with the trend) - More risk with better setups Refactor

Predicting

6/20 15s interval EMA - ZEC, XLM works well - BTC doesn't work as well (lower ratio?) - EXCH sporadic (it's low vol, maybe steer clear anyway)

Unscheduled rebalances

Make this automatic, poll every 1m to refresh basket/NAV. Figure out rebalance price. For tokens that are close to rebalance price, refetch every 1s while close. Buy when very close to rebalance price, with take profits and tight stop-loss. Send SMS after.

Backtest Results

Balls to the wall (50-100x leverage)

  • $1.5m buying, > 15% ratio, EMA Cross in, 0.75%TRL out
-- BY WIN PCT
-- consistently around 0.11-0.125
-- SELECT AVG((args->>'minRatio')::FLOAT) FROM backtest_results WHERE all_stats != '[]' AND win_pct > 0.7

-- all win pcts > 0.7 have $1m+ rebal size
-- SELECT AVG((args->>'minRebalanceAmt')::FLOAT) FROM backtest_results WHERE all_stats != '[]' AND win_pct > 0.7

-- at 0.7 win pct, 60% 0.005, 40% 0.0075
-- at 0.9 win pct, 100% 0.0075 (small sample though)
-- SELECT args->>'outSignal' AS signal FROM backtest_results WHERE all_stats != '[]' AND win_pct > 0.7 ORDER BY signal

-- limit random
-- SELECT args->>'limit' AS signal FROM backtest_results WHERE all_stats != '[]' AND win_pct > 0.8 ORDER BY signal


-- BY PROFIT PCT
-- wide range, 0.08-0.1, no pattern
-- SELECT AVG((args->>'minRatio')::FLOAT) FROM backtest_results WHERE all_stats != '[]' AND profit_pct > 0.75

-- profit_pct increaseswith amt, 700000 -> $1m+
-- SELECT AVG((args->>'minRebalanceAmt')::FLOAT) FROM backtest_results WHERE all_stats != '[]' AND profit_pct > 0.3

-- 0.0075 clear winner
-- SELECT args->>'outSignal' AS signal FROM backtest_results WHERE all_stats != '[]' AND profit_pct > 0.3 ORDER BY signal

-- limit 1 clear winner for all profits > 40%
-- SELECT args->>'limit' AS signal FROM backtest_results WHERE all_stats != '[]' AND profit_pct > 0.5 ORDER BY signal


-- robustness across permutations of:
	-- limit 1
	-- minRebalance > 700k
	-- ratio > 0.08 && ratio < 0.12
	-- outSignal of 0.005 or 0.0075
	
-- SELECT * FROM backtest_results WHERE
-- num_trades > 0 
-- AND (args->>'minRatio')::FLOAT > 0.09
-- AND (args->>'limit')::INT = 1
-- AND (args->>'minRebalanceAmt')::FLOAT > 250000
-- AND (
-- 	(args->>'outSignal') = '["TRL0.005"]'
-- 	OR (args->>'outSignal') = '["TRL0.0075"]'
-- )

/** STRATEGY

  • get order book and figure out how much I can buy without moving market too much. Some of these coins only trade single-digit millions in 24h

    • not a big deal on the buy, as limit means it will get only partially filled, but on the sell it could get ugly

  • size position and stop according to 24h volatility

    • probably too slow to calculate this on the fly, would have to have a recurring task, storage solution

  • size position according to reliability of signal

    • based on twitter account, market size (need to get more data/a feel for this)

    High conviction bets: KALEO + small markets + 'send'

Thought: Even if a tweet doesn't move the market per se, it's likely to be excited about the right things at approximately the right times...

TODO

  • when updating tweets, only add tweets older than an hour, so we have complete +-60 min data for it
  • make 0th min round up (so after the tweet). Perhaps round up +1 interval if in last 10 seconds of period (so the range of buys would be 10s-25s after)?
  • use OHLC to estimate P/L
  • on update, save most recent account info, etc to database?
  • save marketslist to DB
  • make package.json scripts for updating tweet data, etc.
  • "if the tik tokers can send literal dog sh*t, we can send $zec" caught the phrase send lit, but not send $zec

STRATEGY INVERSION start with amt to risk (based on scale) - let riskAmt = 100 get ask - let bid = ask + 0.0001 calculate recent volatility (eg: last hour, in 1 min intervals) - let mean = (float) mean of the 60 intervals - let sqDeviations = ([floats]) square(mean - deviation) - let variance = (float) sum(sqDeviations) / 60 - let periodVol = sqrt(variance) IE: assuming std distribution, 2/3 of returns in 1 period will fall +/- periodVol/2 - let 5periodVol = sqrt(5) * periodVol set % trailing stop based on volatility - let trlStop = periodVol / 2 (or something like that) quantity = bid * riskAmt / trlStop

THINGS TO ACTUALLY BUY $RUNE, $ALPHA - RookieXBT Other than $RUNE which is by far my biggest holding (I buy more everyday regardless of price) — I’ve been holding onto a sizeable amount of $ALPHA for some time And with that being said, I’ve never felt more comfy than I do now. This summer is gonna be good.

IMPORTANT:
Check bid/ask spread, don't play the game if greater than...0.05%?
 */

// relVolChange 0-5
// plp P/L Percent, as a decimal

/**
 * TESTING
 * ensure markets still being picked up correctly
 * elon/jack tweets (10s probably too slow to pick up the big moves)
 *      - can fetch up to every 3 seconds, but that's still 850k/month, over the total project cap of 500k
 */
let trades = [
    {
        tweet: "I am pretty irresponsibly long $CRV here, this long term chart.. ooh...",
        market: 'CRV-PERP',
        buy: 3.7291,
        sell: 0,
        plp: 0,
        volChange: 0,
        notes: ``,
    },
    {
        tweet: "@ShardiB2 - RT @trader1sz: $CRV 🤝 https://t.co/YlCyoqFnzd",
        market: 'CRV-PERP',
        buy: 3.6385,
        sell: 0,
        plp: 0,
        volChange: 0,
        notes: ``,
    },
    {
        tweet: "@ShardiB2 - @solstarterorg is coming.....  one of the better projects I invested in.  If you believe in $SOL than you should probably check out the medium piece here... https://t.co/eDbcffOngk",
        market: 'SOL-PERP',
        buy: 43.7661,
        sell: 43.5424,
        plp: -0.005,
        volChange: 0,
        notes: `Obviously a mistake to buy this, need to ignore the nonsense tweets.`,
    },
    {
        tweet: "@ShardiB2 - @Off_Erwan seriously, I am not even sure what you are talking about, I have not sold my core spot position ever, I have like 6 accounts I trade and if I remember correctly, I CRUSHED that $BAND trade.  Honestly, I am like 90 trades past that.",
        market: 'BAND-PERP',
        notes: `Just sold this one right away, should filter stuff like this out as much as possible. No change in vol/px action though.`,
    },
    {
        tweet: "@CryptoKaleo - As expected, $OMG has been absolutely on fire the past couple of days. https://t.co/in8Yxaf4V",
        market: 'OMG-PERP',
        buy: 12.812,
        sell: 12.7535,
        plp: -0.0045,
        volChange: 4,
        notes: `Huge spike, big gap between fills. Caught it in the middle of the candle.`,
    },
    {
        tweet: "@ShardiB2 - $ATOM!! https://t.co/J5oEcR1KrX https://t.co/sYaxIHUtwf",
        market: 'ATOM-PERP',
        buy: 29.7109,
        sell: 30.759,
        plp: 0.0352, // !!!
        volChange: 1,
        notes: `Small spike, probably didn't contribute much, BUT she wasn't wrong. It kept going up nicely.`,
    },
    {
        tweet: "@ShardiB2 - RT @inmortalcrypto: $ATOM above $100 this summer...missing link",
        market: 'ATOM-PERP',
        buy: 28.921,
        sell: 29.4155,
        plp: 0.017,
        volChange: 0,
        notes: `Small spike, probably didn't contribute much, kept going up nicely.`,
    },
    {
        tweet: "@ShardiB2 - $BNT Buy it if it breaks IMO https://t.co/3CO2db5xXd",
        market: 'BNT-PERP',
        buy: 7.7160,
        sell: 7.851,
        plp: 0.0175,
        volChange: 4,
        notes: 'Nice vol spike, spike started at least 30s after tweet. Pump lasted 1 min, then correction, ranging.',
    },
    {
        tweet: "@ShardiB2 - $CRV S/R  Good entry if not in https://t.co/ca6rMwoLFW",
        market: 'CRV-PERP',
        buy: 3.63855,
        sell: 3.68165,
        plp: 0.011,
        volChange: 3,
        notes: 'Nice vol spike, I got in 100% before price spike.',
    },
    {
        tweet: "@ShardiB2 - Hope you got some $CRV",
        market: 'CRV-PERP',
        buy: 3.74,
        sell: 3.717,
        plp: -0.006,
        volChange: 0,
        notes: 'Nothing changed...weird vol spike & selling 1 min before tweet.'
    },
];

fgv