fibonacci-fit

Find best-fit Fibonacci price retracement levels for an array of prices.

Given a list of prices, this module finds the optimal low/high prices to draw the intermediate Fibonacci levels [0, 0.236, 0.382, 0.5, 0.618, 0.786, 1] such that the MSE between each price and the nearest Fibonacci level, is minimized.

Can also be used to find the best-fit for an arbitrary list of levels rather than the Fibonacci levels.

Installation

npm i fibonacci-fit

Usage

In this example we load some candles from a compressed JSON file, extract a list of prices, then graph the result.

var {fitSlow, fitStochastic, fitStochastic2, encodeIntoNearestLevels, DEFAULT_FIB_LEVELS} = require('fibonacci-fit');

// DEFAULT_FIB_LEVELS = [0, 0.236, 0.382, 0.5, 0.618, 0.786, 1]
// DEFAULT_FIB_LEVELS_EXTENDED = [-0.236, 0, 0.236, 0.382, 0.5, 0.618, 0.786, 1, 1.272];

//fitSlow(prices, _fibLevels = DEFAULT_FIB_LEVELS, minLengthStartToEnd = 10)
// ^^^ slow, inefficient, brute force, find best range based on all prices in the list

//fitStochastic(prices, attempts = prices.length, _fibLevels = DEFAULT_FIB_LEVELS)
// ^^^ stochastic random, choose random prices x[attempts] times and take the best result

//fitStochastic2(prices, stochasticSteps=500, randomStartSteps=20, tempMultiplier = 0.95, _fibLevels = DEFAULT_FIB_LEVELS)
// ^^^ stochastic annealing style, choose randomly x[randomStartSteps] times then x[stochasticSteps] times with temperature decreasing by a factor of tempMultiplier [starting from std/2 of prices]. results are not limited to exactly values of prices like the other functions. 

//both return objects of the same format:
// {
//  start,           //price at the lowest fibonacci level
//  end,             //price at the highest fibonacci level
//  mse,             //mean squared error of price vs nearest fibonacci level
//  fibonacciLevels, // the list of fibonacci levels that we input
//  priceLevels      //the list of prices at each fibonacci level
// }

//also: 
//encodeIntoNearestLevels(prices, doReturnString = false, windowSizeFit=10, minFitWidthFib=1)
//^^^ convert list of prices into list of indices representing 
//   the index of the nearest fib-line for the price at that point, 
//   given fib-lines fitted over the previous windowSizeFit prices, 
//   in a rolling fashion. Result length is same as input.
//
// - set doReturnString to true to get a string like "aabbac..." 
//   [where a=0, b=1, etc] instead of an array of indices [ints].
//
// - note that the first ${windowSizeFit} elements will use smaller windows,
//   and so may be unreliable / random ...

//loading last 100 candles from a dataset... {open, high, low, close, volume}
var candlesFull = require('jsonfile-compressed-brotli').readFileSync('./AAPL.json');
candlesFull=candlesFull.slice(-100);

//extract a list of just closing prices
var prices = candlesFull.map(c=>c.close);
//get the best-fit fibonacci levels   
var fibResults = fitSlow(prices); //or fitStochastic(prices);

var {drawChartForCandles,saveChartForCandles} = require('ohlc-chart-simple');

var config = {
    lines:
        fibResult.priceLevels.map(function(price){
            return {startPrice: price, endPrice: price, startIndex:1, endIndex: 100, color: [0,0,255], thickness:0}
        }),
    w: 512,
    h: 350,
    profileBucketsWidth: 0,
    volumeBarsHeight: 16,
    bgColor: [255,255,255],
    title: "AAPL",
    filename: `./candlestick-chart.png`,
}

saveChartForCandles(candlesFull, config);

//"quantize" prices into index-of-nearest-levels, rolling:

var windowSize = 10; //back-fitting window, default 10, try higher values 
var minFitWidthFib = 1; //default 1
var doReturnString = true; //default false 
console.log(encodeIntoNearestLevels(prices, doReturnString, windowSize, minFitWidthFib));

//aacgdgggggfggfebcaaggggbcdcgggfeffeefggcdaacaeagagfggggfggfggfgfeeddbcecgggeceffggggfggeecagfbaafeaa

doReturnString = false;
console.log(encodeIntoNearestLevels(prices, doReturnString, windowSize, minFitWidthFib));

// [
//     0, 0, 2, 6, 3, 6, 6, 6, 6, 6, 5, 6,
//     6, 5, 4, 1, 2, 0, 0, 6, 6, 6, 6, 1, ...

//notice that these are NOT the candle-by-candle nearest line indices corresponding to the shown graph, 
//as this function is calculated per-candle in a rolling fashion only using 'earlier' data

Resulting plot of best-fit Fibonacci price levels:

See Also

• ohlc-chart-simple - simple OHLCV chart for node.js