Kalman Velocity Strategy [NovaLens]

Kalman velocity trend strategy for BTC/USD daily charts on Coinbase, long only. It buys when the velocity of an adaptive Kalman filter turns up on a confirmed daily close inside an uptrend and a calm market and it rides the move with an ATR trailing stop.
The settings were tuned on data up to 31 December 2025. Everything from 1 January 2026 on is unseen data the tuning never used, and on it the strategy made +16.44% so far while BTC is down -2%.
AT A GLANCE: STRATEGY VS HOLDING BTC
It was profitable in every calendar year from 2018 to 2026 so far. It finished ahead of holding BTC in the 4 years BTC fell and behind in the 5 years it rose; over the whole period it made more than holding BTC, with a drawdown of 23% against 82%.
From 1 January 2018 (first trade 2018-04-13) to 2026-10-05; chart COINBASE:BTCUSD daily, history from 2014-12-01; Tester run 2026-10-06, Properties below, training years included. Trading starts on 1 January 2018; the settings were tuned on data from 17 August 2017 to 31 December 2025; earlier Coinbase candles are only warm-up. Past results, not a forecast.
Sharpe and Sortino: TradingView's method (monthly returns less a 2% yearly cash rate), the strategy's from the Tester, BTC's from month-end closes from January 2018. Bases differ.
YEAR BY YEAR
Tester closed profit after costs by exit year, in % of equity at the year start; BTC from first daily open to last close:
HOW IT WAS BUILT
It starts from the velocity turn of our open-source Kalman Velocity Ribbon. On its own, that turn traded often and did not beat holding BTC. Three rules were added on top: a trend filter, a calm filter and an ATR trailing stop that replaced the exit on the opposite turn.
The settings were tuned on Binance BTCUSDT daily data from 2017 through 2025, about 15,500 settings tested. 2026 was kept apart as unseen data. The Coinbase years before 2017 were never used in tuning.
WHAT EACH RULE DOES, AND WHY
HOW TO READ A TRADE ON THE CHART
Trade Story, the default Chart Style, prints each closed trade's result in percent under its exit candle: green for a gain after fees, red for a loss. Hover it for the entry and exit reasons, dates, prices and candles held. Tick "Show trade boxes" for a shaded zone per trade.
The buy and exit marks are TradingView's own order marks at fill prices; untick "Trades on chart" in the Style tab for results only. The stepped orange line is the open trade's trailing stop, the quiet line the 35-day trend average ("Show trailing stop", "Show trend average"). The panel shows the position, trend strength, trade age, the stop, its distance, net profit, 2026 profit, max drawdown, win rate and trades.
Signal Map tints the background while all three entry conditions hold and marks entries and exits with triangles (tooltip: the reason); Clean Pro shows a thin grey stop, the triangles and a result chip per exit.
HOW TO SET THE ALERT
Create an alert on this strategy, keep "Order fills and alert() function calls" and put this in Message: {{strategy.order.alert_message}} Size {{strategy.order.contracts}}; fill {{strategy.order.price}}. Each fill says buy or sell, why (exits name the reason), size and fill price. In a trade, a heads-up says when the stop moved up.
WHAT THE DEFAULT REPORT SHOWS
The average trade lasts about 8 candles; the average winner made 11.01% of the position and the average loser lost 3.97%. It was in a trade about one day in six.
The default makes 71 trades over almost 9 years on COINBASE:BTCUSD 1D because it only buys a Kalman velocity turn inside an uptrend and a calm market, so each full year's result rests on 4 to 11 trades and the sample is small.
BTC's drawdown is its daily low against the highest earlier close since 1 January 2018, Jan 2018 to Dec 2018. The two drawdowns are measured differently; the comparison is descriptive.
UNSEEN DATA: 2026
The settings were tuned on data up to the end of 2025; 2026 was not used. The dashed line on the chart marks 1 January 2026: training on its left, unseen data on its right. 2026 has six trades, and the trade entered on 2026-08-19 accounts for most of the gain.
The Tester's 2026 profit is +16.44%; in our simulation the 2026-08-19 trade made 21,304.88 of the 21,800.06 USD gained. Holding BTC from the first 2026 open to the 2026-10-05 close lost 2.00%.
HOW THESE NUMBERS WERE MADE
AT A GLANCE, the year split and the 2026 profit come from the TradingView Strategy Tester on this chart, Properties below. Trade averages and stop distances come from our own simulation of the same rules on the same candles: the same 71 trades, net profit within 0.3% of the Tester. BTC's own changes come from the chart's daily candles. The trailing stop is a stop order: if a candle opens below the stop, TradingView fills the exit at that open, so a loss can be larger than the stop distance (none of the 71 trades here exited that way). The stability and six-coin counts below are from our own test runs on Binance data, not Tester reports.
HONEST LIMITATIONS
WHY ACCESS IS LIMITED
The added rule set (trend filter, volatility filter, ATR trailing stop, tuned together with the filter speed on Binance BTCUSDT daily) and the costed trade report are the basis for limited access. The Kalman filter and its velocity-turn signal are those of the open-source Kalman Velocity Ribbon indicator.
PROPERTIES AND ACCESS
10,000 USD capital, 99% of equity per trade, commission 0.10% per order (a typical exchange fee plus slippage), Slippage field 1 tick, margin 100%, no pyramiding, COINBASE:BTCUSD daily, standard candles. Risk per trade: the first trailing stop sat a median 4.64% below the signal close and 7.50% or less in nine trades of ten, so a stopped trade typically risks about 4.6% of equity, before any gap.
Invite-only and closed source; access is granted by hand on request. Look, Theme, Panel, Chart Style, Show trade boxes and Show unseen-data line change only the display, never the orders.
The settings were tuned on data up to 31 December 2025. Everything from 1 January 2026 on is unseen data the tuning never used, and on it the strategy made +16.44% so far while BTC is down -2%.
AT A GLANCE: STRATEGY VS HOLDING BTC
It was profitable in every calendar year from 2018 to 2026 so far. It finished ahead of holding BTC in the 4 years BTC fell and behind in the 5 years it rose; over the whole period it made more than holding BTC, with a drawdown of 23% against 82%.
From 1 January 2018 (first trade 2018-04-13) to 2026-10-05; chart COINBASE:BTCUSD daily, history from 2014-12-01; Tester run 2026-10-06, Properties below, training years included. Trading starts on 1 January 2018; the settings were tuned on data from 17 August 2017 to 31 December 2025; earlier Coinbase candles are only warm-up. Past results, not a forecast.
- Net profit: strategy +1,438.65% after costs, holding BTC +518.54% (from 1 January 2018; Tester +988.53% starts at the first trade)
- Max drawdown: strategy 22.88%, holding BTC 81.70%
- Return/drawdown: strategy 62.88, holding BTC 6.35
- Sharpe ratio: strategy 0.33, holding BTC 0.17
- Sortino ratio: strategy 1.13, holding BTC 0.32
- Profitable calendar years: strategy 9 of 9, BTC 5 of 9
- Years ahead of BTC: 4 of 9: 2018, 2022, 2025, 2026 so far; behind in 2019, 2020, 2021, 2023, 2024
- Strategy only: profit factor 2.98, win rate 57.7%, 71 trades
Sharpe and Sortino: TradingView's method (monthly returns less a 2% yearly cash rate), the strategy's from the Tester, BTC's from month-end closes from January 2018. Bases differ.
YEAR BY YEAR
Tester closed profit after costs by exit year, in % of equity at the year start; BTC from first daily open to last close:
- 2018: 4 trades, +17.75%; BTC -73.37%
- 2019: 10 trades, +89.96%; BTC +94.09%
- 2020: 11 trades, +79.29%; BTC +304.57%
- 2021: 7 trades, +42.12%; BTC +59.40%
- 2022: 5 trades, +12.42%; BTC -64.23%
- 2023: 10 trades, +26.56%; BTC +155.80%
- 2024: 9 trades, +39.53%; BTC +120.76%
- 2025: 9 trades, +16.79%; BTC -6.27%
- 2026 to 2026-10-05: 6 trades, +16.44%; BTC -2.00%
HOW IT WAS BUILT
It starts from the velocity turn of our open-source Kalman Velocity Ribbon. On its own, that turn traded often and did not beat holding BTC. Three rules were added on top: a trend filter, a calm filter and an ATR trailing stop that replaced the exit on the opposite turn.
The settings were tuned on Binance BTCUSDT daily data from 2017 through 2025, about 15,500 settings tested. 2026 was kept apart as unseen data. The Coinbase years before 2017 were never used in tuning.
WHAT EACH RULE DOES, AND WHY
- Entry turn: a Kalman filter tracks a smoothed price and its speed (velocity) on every closed candle; a new price counts for less when ATR runs above its average. A turn is the velocity moving from zero or below to above zero on a closed candle. Why: it marks where the smoothed price stops falling and starts rising.
- Trend filter: a long is taken only when the close is above its 35-day simple moving average. Why: it keeps the strategy out of falling markets, where an upturn can be a short bounce.
- Calm filter: no new long while the 14-day ATR is 8% of the close or more; an open trade is not affected. Why: with the stop 1.25 ATR away, one stopped trade would cost about 10% of the account or more.
- Entry fill: when the turn and both filters agree and no trade is open, it buys at the next candle's open.
- Trailing stop: placed 1.25 times ATR(14) below the signal close together with the entry order, so it already works on the fill candle. After each daily close it moves up to 1.25 ATR below that close when this is higher, and it never moves down. Why: it lets a trend run and exits on the first real pullback, sooner than a velocity turn down.
- Position size, 99% of equity: each trade uses the whole account except about 1%, kept back so the commission on entry and exit is always covered; at a full 100% the Tester skipped fills and logged margin calls.
- Other inputs: Exit on Velocity Turn, off by default, also closes on a turn down; a percent Stop Loss and Take Profit are off by default. No trades during the 54-candle warm-up.
HOW TO READ A TRADE ON THE CHART
Trade Story, the default Chart Style, prints each closed trade's result in percent under its exit candle: green for a gain after fees, red for a loss. Hover it for the entry and exit reasons, dates, prices and candles held. Tick "Show trade boxes" for a shaded zone per trade.
The buy and exit marks are TradingView's own order marks at fill prices; untick "Trades on chart" in the Style tab for results only. The stepped orange line is the open trade's trailing stop, the quiet line the 35-day trend average ("Show trailing stop", "Show trend average"). The panel shows the position, trend strength, trade age, the stop, its distance, net profit, 2026 profit, max drawdown, win rate and trades.
Signal Map tints the background while all three entry conditions hold and marks entries and exits with triangles (tooltip: the reason); Clean Pro shows a thin grey stop, the triangles and a result chip per exit.
HOW TO SET THE ALERT
Create an alert on this strategy, keep "Order fills and alert() function calls" and put this in Message: {{strategy.order.alert_message}} Size {{strategy.order.contracts}}; fill {{strategy.order.price}}. Each fill says buy or sell, why (exits name the reason), size and fill price. In a trade, a heads-up says when the stop moved up.
WHAT THE DEFAULT REPORT SHOWS
The average trade lasts about 8 candles; the average winner made 11.01% of the position and the average loser lost 3.97%. It was in a trade about one day in six.
The default makes 71 trades over almost 9 years on COINBASE:BTCUSD 1D because it only buys a Kalman velocity turn inside an uptrend and a calm market, so each full year's result rests on 4 to 11 trades and the sample is small.
BTC's drawdown is its daily low against the highest earlier close since 1 January 2018, Jan 2018 to Dec 2018. The two drawdowns are measured differently; the comparison is descriptive.
UNSEEN DATA: 2026
The settings were tuned on data up to the end of 2025; 2026 was not used. The dashed line on the chart marks 1 January 2026: training on its left, unseen data on its right. 2026 has six trades, and the trade entered on 2026-08-19 accounts for most of the gain.
The Tester's 2026 profit is +16.44%; in our simulation the 2026-08-19 trade made 21,304.88 of the 21,800.06 USD gained. Holding BTC from the first 2026 open to the 2026-10-05 close lost 2.00%.
HOW THESE NUMBERS WERE MADE
AT A GLANCE, the year split and the 2026 profit come from the TradingView Strategy Tester on this chart, Properties below. Trade averages and stop distances come from our own simulation of the same rules on the same candles: the same 71 trades, net profit within 0.3% of the Tester. BTC's own changes come from the chart's daily candles. The trailing stop is a stop order: if a candle opens below the stop, TradingView fills the exit at that open, so a loss can be larger than the stop distance (none of the 71 trades here exited that way). The stability and six-coin counts below are from our own test runs on Binance data, not Tester reports.
HONEST LIMITATIONS
- Thin years: 2018 rests on 4 trades and 2022 on 5.
- Stability: each tuned input moved 20% up and down, one at a time, at 0.10% cost: every year 2018 to 2025 stayed positive in all ten moves, but the thinnest (Max Volatility % at 6.4) left 2018 just above zero, and at 0.15% per order slightly negative.
- The calm filter has not been active since 2022: the ATR reached 8% of the close on 120 days in 2018, 87 in 2021, 16 in 2022 and on none since 2023.
- Selection bias: tuned on Binance 2017-2025, the best of about 15,500 tries.
- Starting earlier changes the comparison: from 1 January 2017 holding BTC made +8,709%, far more than the strategy's +2,640% (our simulation), as BTC rose about 4.5 times before 17 August 2017.
- The search setup was finalized in 2026 with the full history available; run on Binance data up to the end of 2025 only, the same search returns these settings.
- Fitted to BTC only. The same settings on ETHUSDT daily had three losing years, so no other market or timeframe is claimed. Across six coins (BTC, ETH, SOL, BNB, XRP, ADA) on 4h, 12h and daily bars, 4 of 18 combinations had net profit above zero, profit factor at least 1.3 and at most one losing year: the tuned BTC daily case and 3 of the other 17.
- Lag and short holds: the velocity turns after price does, so entries come after the low, and the stop cuts long trends into shorter trades.
- Funding payments on perpetual futures and differences between exchanges and fee tiers are not modelled.
- Changing a rule input leaves the tuned setting; the report then no longer describes it. Heikin Ashi, Renko and similar charts send no orders.
WHY ACCESS IS LIMITED
The added rule set (trend filter, volatility filter, ATR trailing stop, tuned together with the filter speed on Binance BTCUSDT daily) and the costed trade report are the basis for limited access. The Kalman filter and its velocity-turn signal are those of the open-source Kalman Velocity Ribbon indicator.
PROPERTIES AND ACCESS
10,000 USD capital, 99% of equity per trade, commission 0.10% per order (a typical exchange fee plus slippage), Slippage field 1 tick, margin 100%, no pyramiding, COINBASE:BTCUSD daily, standard candles. Risk per trade: the first trailing stop sat a median 4.64% below the signal close and 7.50% or less in nine trades of ten, so a stopped trade typically risks about 4.6% of equity, before any gap.
Invite-only and closed source; access is granted by hand on request. Look, Theme, Panel, Chart Style, Show trade boxes and Show unseen-data line change only the display, never the orders.
Systematic indicators and strategies from real quant research, designed for a cleaner market view.
🟢 Join our research list: novalens.io
Not financial advice. Always do your own research.
🟢 Join our research list: novalens.io
Not financial advice. Always do your own research.
Disclaimer
The information and publications are not meant to be, and do not constitute, financial, investment, trading, or other types of advice or recommendations supplied or endorsed by TradingView. Read more in the Terms of Use.
Systematic indicators and strategies from real quant research, designed for a cleaner market view.
🟢 Join our research list: novalens.io
Not financial advice. Always do your own research.
🟢 Join our research list: novalens.io
Not financial advice. Always do your own research.
Disclaimer
The information and publications are not meant to be, and do not constitute, financial, investment, trading, or other types of advice or recommendations supplied or endorsed by TradingView. Read more in the Terms of Use.