Relative strength index (RSI)

The code

The function returns one RSI value per day. The smoothing is Wilder's own: after the first plain average, each new day counts for one fourteenth and the old average for the rest. Many libraries differ slightly in how they start the series, which is why two tools can disagree for the first few weeks of a file.

The second half of the script uses the reading in a simple rule, only to show how a rule built on an indicator is tested. The thresholds are the textbook ones.

Data the script reads: invented-prices.csv. Invented values, never a real market. Save it in one folder and run the script from that folder.

The file that was run: rsi.py

# The relative strength index (RSI), computed from scratch with no library.
# Prices come from invented-prices.csv: a random walk made by a script, never a real market.
# The result therefore says nothing about any real market. Read it for the method.
import csv

closes = []
with open("invented-prices.csv", newline="") as f:
    for row in csv.DictReader(f):
        closes.append(float(row["close"]))


def rsi(prices, period):
    """Return a list the same length as `prices`. Positions with too little history hold None.

    The value at position i uses prices up to and including i, and nothing after it.
    """
    values = [None] * len(prices)
    gains, losses = [], []
    for i in range(1, len(prices)):
        change = prices[i] - prices[i - 1]
        gains.append(max(change, 0.0))
        losses.append(max(-change, 0.0))

    # Start with a plain average of the first `period` gains and losses.
    avg_gain = sum(gains[:period]) / period
    avg_loss = sum(losses[:period]) / period

    for i in range(period, len(prices)):
        if i > period:
            # Wilder's smoothing: keep most of the old average, blend in one new day.
            avg_gain = (avg_gain * (period - 1) + gains[i - 1]) / period
            avg_loss = (avg_loss * (period - 1) + losses[i - 1]) / period
        if avg_loss == 0:
            values[i] = 100.0
        else:
            strength = avg_gain / avg_loss
            values[i] = 100 - 100 / (1 + strength)
    return values


# Example values, picked only because they are easy to follow. They are not a suggestion.
PERIOD = 14
LOW, MIDDLE, HIGH = 30, 50, 70
COST = 0.10          # points paid on every completed trade, a stand in for the spread

index = rsi(closes, PERIOD)
known = [v for v in index if v is not None]
print("Last five RSI values:", [round(v, 1) for v in known[-5:]])
print("Lowest and highest:", round(min(known), 1), "and", round(max(known), 1))
print(f"Days below {LOW}:", len([v for v in known if v < LOW]), "of", len(known))
print(f"Days above {HIGH}:", len([v for v in known if v > HIGH]), "of", len(known))

# A rule built on it, to show how one is tested: in when RSI was under LOW, out when back over MIDDLE.
holding = False
entry = 0.0
results = []
START = PERIOD + 1

for day in range(START, len(closes)):
    reading = index[day - 1]     # yesterday's RSI, which only knows prices up to yesterday
    price = closes[day]
    if not holding and reading < LOW:
        holding, entry = True, price
    elif holding and reading > MIDDLE:
        results.append(price - entry - COST)
        holding = False

if holding:
    results.append(closes[-1] - entry - COST)

print("Trades:", len(results))
print("Trades that ended above zero:", len([r for r in results if r > 0]))
print("Total points after costs:", round(sum(results), 2))
print("Simply holding over the same days:", round(closes[-1] - closes[START] - COST, 2))

What the run printed

Last five RSI values: [47.6, 46.9, 44.8, 44.5, 41.7]
Lowest and highest: 17.4 and 73.7
Days below 30: 38 of 736
Days above 70: 12 of 736
Trades: 4
Trades that ended above zero: 2
Total points after costs: -1.93
Simply holding over the same days: -9.96

Output of the run on 9 Oct 2026, copied as printed. Every input to this run was invented, so the result says nothing about any real market.

How to read it

The first four lines describe the indicator itself. On this file the reading spent most of its time between the two thresholds, dipped under 30 on 38 days and went over 70 on only 12. Days under the threshold come in clusters, which is why 38 days produced so few trades.

The rule made four trades in 736 days. Nothing can be concluded from four trades, whichever way they went. The page on how a backtest lies shows how often pure chance produces a good looking record from a handful of trades.

The total is below zero and the comparison with holding reflects being out of a falling file most of the time. The prices are a random walk, so a low reading here does not make a rise more likely. It says nothing about any real market.