What's in a genome

A genome is the settings card behind each strategy. Here is every setting, in plain words.

The idea

Every strategy in this project is a genome: a fixed card of 21 settings. Some are numbers (how many stocks to hold, how often to trade). Some are choices from a short list (which ranking rule to use). A bot reads its card and follows it. That is all a bot is. All trading here uses paper money, never real money.

Evolution only changes the values on the card. It copies a card, nudges some numbers, flips some choices, or mixes two cards. It never writes code. The program that turns a card into trades is fixed and was written by people.

How a genome becomes trades each day

Read it top to bottom. Costs are small: 0.02% per trade for funds, 0.05% for stocks in the S&P 500 and Nasdaq-100, 0.20% for small-cap stocks.

The bot only sees prices through today's close. Its orders fill at tomorrow's open, so it never trades at a price it could not have known.
Steps 3 to 6 run on rebalance days, or on the day the market filter flips. Exit rules are checked every day at the close.
The kill switch is not a setting. If an account falls 35% below its highest value, the bot is switched off. Position caps and a limit of 40 new trades a day are also fixed.
The bot cannot borrow or sell short. It can only buy with cash it has.

Core (1 setting)

How much money simply owns the index fund.

Setting	Range or options	Unit	What it does	High value	Low value
Index-fund core `core.pct`	0 to 100	percent of money	The share of paper money kept in the arena's index fund (SPY for S&P 500, QQQ for Nasdaq-100, IWM for small caps). The rest is the strategy money. At 100 the strategy settings do nothing.	Closer to simply owning the index. Results track the market more and the strategy less.	The strategy drives the result. At 0 there is no index fund.

Market filter (regime) (4 settings)

An optional switch that moves money to a safe place when the market looks weak.

Setting	Range or options	Unit	What it does	High value	Low value
Market filter `regime.enabled`	yes or no	yes or no	Switches the market filter on. Each day the bot compares the index fund's closing price with its recent average. If the price is below the average, the market counts as weak and money moves to the safe haven. When the filter flips, the bot rebalances at once instead of waiting for the next rebalance day.	Yes: the filter is active.	No: the filter is ignored, so its average length and the core setting below do nothing.
Filter length `regime.ma_days`	20 to 250	trading days	How many trading days go into the index's average price. A moving average is the plain average of the last N closing prices.	A slow filter. It ignores short dips and reacts late.	A fast filter. It reacts quickly but flips more often, so it trades more.
Safe haven `regime.safe_haven`	cash, bonds, gold	a choice	Where money goes when the market looks weak: cash (BIL, a cash-like fund), bonds (IEF) or gold (GLD). The same choice is used by the hold_safe rule, even when the filter is off.	n/a (a choice)	n/a (a choice)
Filter moves the core `regime.apply_to_core`	yes or no	yes or no	When the market looks weak, also move the index-fund core to the safe haven. If no, only the strategy money moves and the core stays in the index.	Yes: everything moves to the safe haven.	No: the core stays invested through weak markets.

Selection (9 settings)

What the strategy money buys, and how candidates are ranked.

Setting	Range or options	Unit	What it does	High value	Low value
Pool `selection.pool`	stocks, sectors, assets	a choice	What the strategy money can buy. stocks: the arena's stocks that were index members on that day. sectors: the 11 sector funds. assets: SPY, QQQ, IWM, EFA, EEM, IEF, TLT, GLD and BIL (stock, bond, gold and cash funds). One stock may hold at most 20% of the account, one sector fund 40%.	n/a (a choice)	n/a (a choice)
Ranking rule `selection.rule`	momentum, trend, pullback, low_volatility, reversal, hold_safe	a choice	How candidates are ranked. momentum: biggest gain over the lookback, and the gain must be above zero. trend: furthest above the slow average. pullback: above the slow average and up over the lookback, ranked by how far the price sits below the fast average. low_volatility: the smoothest day-to-day prices. reversal: biggest loss over the lookback. hold_safe: no ranking; the strategy money sits in the safe haven.	n/a (a choice)	n/a (a choice)
Lookback `selection.lookback_days`	3 to 300	trading days	How far back the rule looks to measure gain or loss (momentum, pullback, reversal) or the size of daily swings (low_volatility). The trend and hold_safe rules ignore it. 21 trading days is about a month, 252 about a year.	Looks at slow, long trends.	Looks at recent moves, which are noisier.
Skip recent days `selection.skip_days`	0 to 30	trading days	Ignores the most recent N days when measuring the gain or loss. Textbook 12-1 momentum skips the last month (21 days). Used by momentum, pullback and reversal; low_volatility and trend do not skip.	Skips more recent history.	0 means nothing is skipped.
Fast average `selection.sma_fast`	3 to 100	trading days	The short moving average. Only the pullback rule uses it: it wants stocks that have dipped below this average. It must be smaller than the slow average.	A less sensitive 'dip' measure.	A very sensitive 'dip' measure.
Slow average `selection.sma_slow`	10 to 250	trading days	The long moving average. Used by the trend rule, the pullback rule and the trend filter. Every genome also waits until it has this many days of history before it can pick anything.	A slower definition of an uptrend.	A quicker definition of an uptrend.
Trend filter `selection.trend_filter`	yes or no	yes or no	Only buy candidates whose price is above their slow average, whatever the ranking rule says.	Yes: fewer, stronger-looking candidates.	No: no extra check.
Volume filter `selection.min_volume_rank`	0 to 0.95	fraction of pool	Stocks pool only. Ranks stocks by average daily trading volume over the last 20 days and skips those below this fraction. 0.5 skips the least-traded half.	Only heavily traded stocks qualify.	0 means no volume filter.
How many to hold `selection.top_n`	1 to 25	positions	How many of the best-ranked candidates to hold. Fewer may qualify, and then it holds fewer. With stocks, no single stock gets more than about 19% (a 20% cap with a little headroom), so below five names some money stays in cash.	More names, spread more thinly.	Fewer names, bigger bets.

Rebalance (1 setting)

How often the bot trades toward its picks.

Setting	Range or options	Unit	What it does	High value	Low value
Rebalance every `rebalance.every_days`	1 to 63	trading days	Trading days between rebalances, when the bot recomputes its picks and trades toward them. 5 is about weekly and 21 about monthly. Exits and market-filter flips can still trade in between. Trades smaller than 1% of the account are skipped.	Fewer trades and lower costs, but slower to react.	More trades, more costs, faster reactions.

Sizing (2 settings)

How the strategy money is split among the picks.

Setting	Range or options	Unit	What it does	High value	Low value
Sizing `sizing.method`	equal, inverse_vol	a choice	equal: every pick gets the same amount. inverse_vol: each pick gets an amount in proportion to one divided by its recent volatility, so calmer holdings get more.	n/a (a choice)	n/a (a choice)
Volatility window `sizing.vol_lookback_days`	10 to 252	trading days	Days of price history used to measure volatility for inverse_vol sizing. Ignored with equal sizing, though the bot still waits for this much history before it starts.	A steadier, slower volatility estimate.	A jumpier estimate that follows recent swings.

Exits (4 settings)

Rules that sell a stock before the next rebalance. Each is off when set to 0. They apply to stocks and funds the strategy picked, never to the index-fund core or the safe haven.

Setting	Range or options	Unit	What it does	High value	Low value
Stop-loss `exit.stop_loss_pct`	0 to 50 (0 = off)	percent	Sells a stock that closes this far below the price paid. The order fills at the next open. The stock can be bought again at a later rebalance.	A loose stop: it takes bigger losses before selling.	A tight stop: it sells on small dips. 0 means no stop-loss.
Trailing stop `exit.trailing_stop_pct`	0 to 50 (0 = off)	percent	Sells a stock that closes this far below its highest close since it was bought.	A loose stop that gives back more gains.	A tight stop. 0 means off.
Take-profit `exit.take_profit_pct`	0 to 300 (0 = off)	percent	Sells a stock that closes this far above the price paid, locking in the gain.	Lets winners run further.	Sells winners early. 0 means off.
Max hold `exit.max_hold_days`	0 to 500 (0 = off)	trading days	Sells a stock after it has been held this many trading days.	Holds for a long time.	Forces turnover. 0 means off.

What evolution can and can't change

Evolution can

Change any number, within its fixed range.
Switch between the listed options, such as ranking rules or safe havens.
Turn an exit rule on or off by setting it to 0.
Mix two parent cards, taking each whole group from one parent or the other.

Evolution cannot

Write or change code, or invent a new ranking rule.
Change the position caps, trading costs, the 35% kill switch or the next-open fills.
Choose the market. The arena (S&P 500, Nasdaq-100 or small caps) is set by where the bot is evolved.
See future prices, borrow money or sell short.