Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/backtest-kit/uzse-backtest-app/llms.txt

Use this file to discover all available pages before exploring further.

build-candles.ts is the final step in the data pipeline. It reads every trade for a given ISIN from the trade-results collection, aggregates them into 1-minute buckets, forward-fills every gap (intraday minutes with no activity and entire non-trading days), and then fans the filled 1-minute series out into 10 additional higher timeframes — writing all resulting candles to the candle-items collection.

​
Usage

npx tsx scripts/build-candles.ts [symbol] [isin]

​
Parameters

​
symbol
string
default:"HMKB"
The display ticker symbol stored on each candle document, used by the backtest-kit editor to query candles.
Example: HMKB
​
isin
string
default:"UZ7011340005"
The ISIN code used to look up trades in the trade-results collection.
Example: UZ7011340005

​
Example

npx tsx scripts/build-candles.ts HMKB UZ7011340005

​
Supported Timeframes

The script generates candles for all 11 timeframes in a single pass:
IntervalMinutes
1m1
3m3
5m5
15m15
30m30
1h60
2h120
4h240
6h360
8h480
1d1440

​
Algorithm

The candle-building process follows five numbered steps, each building on the output of the previous.
1

Aggregating Real Minutes

Each trade is placed into a 1-minute bucket using floor(time, 1m). Multiple trades within the same minute are combined into a single OHLCV candle:
  • open — price of the first trade in the bucket
  • high / low — maximum / minimum price across all trades
  • close — price of the last trade in the bucket
  • volume — sum of quantity (number of securities, not UZS value)
The script processes trades one calendar day at a time, loading only that day’s records from MongoDB to keep memory usage flat across long histories.
2

Filling Intraday Gaps

A continuous 1-minute series is generated from 00:00 to 23:59 for every day in the range. Minutes that have no real trades are filled with:
  • OHLC = close of the most recent preceding real candle
  • volume = 0
This ensures every timeframe has a contiguous series with no missing bars, as required by the backtest-kit editor.
3

Filling Non-Trading Days

Days on which the exchange was closed (weekends, public holidays) contain no trades at all. All 1,440 minutes of such a day receive:
  • OHLC = close of the last day that had real trades
  • volume = 0
The carry-forward value (lastClose) is tracked across the entire day loop so it propagates correctly through consecutive non-trading days.
4

Aggregating Higher Timeframes

All 11 timeframes (including 1m) are accumulated simultaneously in a single pass over the filled 1-minute series for each day. Each higher-timeframe bucket is keyed by floor(timestamp, N minutes):
  • open — from the first 1m candle of the period
  • high / low — max / min across all 1m candles in the period
  • close — from the last 1m candle of the period
  • volume — sum of volume across all 1m candles in the period
At the end of each day the accumulated documents for all timeframes are batch-inserted into MongoDB in a single insertMany call.
5

Idempotency

The candle-items collection has a compound unique index on { symbol, interval, timestamp }. If a candle for a given symbol, timeframe, and timestamp already exists, the insertMany skips it (using ordered: false) rather than throwing an error. The script tallies both inserted and skipped counters and prints them at the end.

​
MongoDB Collection

Candles are stored in the candle-items collection with the following document shape (from Candle.schema.ts): 
{
  symbol:    string,  // display ticker, e.g. "HMKB"
  interval:  string,  // e.g. "1m", "1h", "1d"
  timestamp: number,  // Unix epoch milliseconds (UTC)
  open:      number,
  high:      number,
  low:       number,
  close:     number,
  volume:    number,
}
The unique index is { symbol: 1, interval: 1, timestamp: 1 }.

​
Progress Output

The script writes a single auto-updating line to stdout per day, showing live progress: 
[47/3012] 2018-04-17  trades:12  real_min:9  candles:15840  inserted:744480  skipped:0
At the end of the full run a summary is printed: 
Done. Inserted: 15163200, skipped (duplicates): 0
Run build-candles.ts only after all the trade data you want is already in MongoDB. Re-running the script at any time is safe — any candles that were already built will be counted as skipped and no duplicates will be created.
If you have a pre-built candle dump (e.g. backtest.candle-items.json), you can bypass the build step entirely and import directly with mongoimport:
mongoimport --db backtest --collection candle-items --file backtest.candle-items.json --jsonArray

Build docs developers (and LLMs) love

Get started for freeTalk to us