Indicators in tradedesk¶

Purpose of this document¶

This guide explains what indicators are within the tradedesk framework, how they are expected to behave, and the mathematical intuition behind the indicators commonly used by strategies.

The goal is not to provide trading advice, but to ensure that indicator usage is:

conceptually correct,
mathematically understood,
and operationally sound when used in both backtests and live execution.

What an indicator is (in tradedesk)¶

In tradedesk, an indicator is a stateful transformation of market data. Indicators:

Consume a stream of prices, candles, or volumes
Maintain internal state across updates
Produce derived values (not trading decisions)

Indicators do not:

Place orders
Know about positions
Encode strategy logic

They are deliberately isolated from execution concerns.

Indicator lifecycle¶

All indicators follow the same conceptual lifecycle:

Construction – parameters are fixed (e.g. lookback periods)
Warmup – insufficient data to produce reliable output
Ready – outputs are mathematically meaningful
Steady-state update – one update per tick or candle

Warmup and readiness¶

Most indicators require a minimum amount of historical data before their output stabilises. During warmup:

Values may be undefined or misleading
Strategies must assume indicators are not ready

In tradedesk, readiness is the responsibility of the strategy or its state container (e.g. EpicState), not the framework itself.

Tick-driven vs candle-driven indicators¶

Indicators differ in how frequently they should be updated:

Tick-driven indicators¶

Updated on every price change
Capture short-term momentum and microstructure
Sensitive to noise

Example: MACD on a fast event stream

Candle-driven indicators¶

Updated once per completed candle
Capture higher-level structure
Less sensitive to noise

Examples: ATR, MFI, Williams %R

A strategy may legitimately combine both, but must do so with a clear understanding of update ordering and lag.

Mathematical foundations¶

This section introduces the indicators commonly used in tradedesk strategies. Each indicator is described in three layers:

Conceptual meaning
Mathematical definition
Practical implementation considerations

Average True Range (ATR)¶

What ATR measures¶

ATR measures price volatility, not direction. It answers the question:

“How much does this market typically move per period?”

ATR is commonly used for:

Position sizing
Stop distance calculation
Volatility regime detection

Mathematical definition¶

For each candle t, the true range (TR) is defined as:

High(t) − Low(t)
|High(t) − Close(t−1)|
|Low(t) − Close(t−1)|

The true range is the maximum of these three values.

ATR is then calculated as a moving average of TR over N periods (typically using an exponential or Wilder-style smoothing).

Interpretation¶

High ATR → volatile market
Low ATR → compressed or ranging market

ATR has no directional bias.

Implementation notes in tradedesk¶

ATR is candle-driven
ATR must be fully warmed up before being used for stops
Strategies should treat a missing or zero ATR as “not ready”

ATR is often used to derive a stop level:

stop_distance = atr * multiplier

The choice of multiplier is a strategy parameter, not an indicator concern.

Keltner Channel¶

Conceptual overview¶

Keltner Channel combines a trend estimate with a volatility envelope.

It uses:

An EMA for the centre line
ATR for the channel width

This makes it useful for:

Breakout filters
Trend-following entry logic
Volatility-aware trailing logic

Mathematical definition¶

For a period N and multiplier M:

Middle = EMA(close, N)
Upper = EMA(close, N) + M × ATR(N)
Lower = EMA(close, N) − M × ATR(N)

Interpretation¶

Price near the upper band suggests strong upside expansion
Price near the lower band suggests strong downside expansion
A narrow channel suggests compression; a widening channel suggests rising volatility

Implementation notes in tradedesk¶

KeltnerChannel is candle-driven
The current implementation uses the same period for both EMA and ATR
update(...) returns a mapping with middle, upper, and lower
The indicator is only ready once both the internal EMA and ATR are ready

As with ATR, strategies should treat None channel values as a warmup condition, not as a trade signal.

Moving Average Convergence Divergence (MACD)¶

Conceptual overview¶

MACD measures momentum by comparing two exponential moving averages (EMAs) of price.

It captures:

Directional bias
Momentum acceleration and deceleration

Mathematical definition¶

Let:

EMA_fast = EMA(price, fast_period)
EMA_slow = EMA(price, slow_period)

Then:

MACD line = EMA_fast − EMA_slow
Signal line = EMA(MACD line, signal_period)
Histogram = MACD line − Signal line

Interpretation¶

Positive histogram → bullish momentum
Negative histogram → bearish momentum
Histogram crossing zero → momentum regime change

Implementation notes in tradedesk¶

MACD can be driven by any consistent price-event stream
Candle closes and tick updates will produce different lag and noise profiles
MACD requires substantial warmup to stabilise

MACD signals are particularly sensitive to:

Update frequency
Price source (mid vs bid/offer)

Money Flow Index (MFI)¶

Conceptual overview¶

MFI is a volume-weighted momentum oscillator.

It attempts to detect:

Accumulation
Distribution
Overbought and oversold conditions

Mathematical definition (high-level)¶

Compute typical price per candle
Multiply by volume to obtain raw money flow
Separate positive and negative flows
Compute the ratio over N periods

The final MFI value is normalised to a 0–100 range.

Interpretation¶

MFI > 80 → overbought
MFI < 20 → oversold

Exact thresholds are strategy-dependent.

Implementation notes in tradedesk¶

MFI is candle-driven
Volume quality matters
MFI is undefined until fully warmed up

Williams %R¶

Conceptual overview¶

Williams %R measures where the current close sits within the recent high–low range.

It answers:

“Is price closing near the top or bottom of its recent range?”

Mathematical definition¶

For a lookback window of N periods:

%R = (HighestHigh − Close) / (HighestHigh − LowestLow) * −100

Values range from −100 to 0.

Interpretation¶

Near 0 → price closing near recent highs
Near −100 → price closing near recent lows

Implementation notes in tradedesk¶

Candle-driven
Sensitive to lookback choice
Often used in conjunction with MFI or trend filters

Indicator interactions¶

Combining indicators introduces non-obvious effects:

Lag compounds across indicators
Agreement can occur after the move
Disagreement may still be informative

Strategies should be designed to tolerate disagreement and delayed confirmation.

Practical considerations¶

Lookback selection¶

Longer lookbacks:

Reduce noise
Increase lag

Shorter lookbacks:

Increase responsiveness
Increase false signals

Timeframe effects¶

Indicators behave differently across timeframes. A strategy should not assume that parameters transfer cleanly from one timeframe to another.

Readiness checks¶

Strategies should explicitly guard against using indicators before they are ready. A missing or unstable indicator value should prevent entries, not merely degrade signal quality.

Exponential Moving Average (EMA)¶

Conceptual overview¶

An Exponential Moving Average (EMA) is a smoothed average of price that places greater weight on more recent observations.

Compared to a simple moving average, an EMA responds more quickly to recent price changes, making it more suitable for momentum-sensitive indicators.

Mathematical definition¶

For a period length N, the smoothing factor α is defined as:

α = 2 / (N + 1)

The EMA is then updated recursively:

EMA(t) = α × Price(t) + (1 − α) × EMA(t−1)

Interpretation¶

EMA tracks the prevailing price direction
Shorter periods increase responsiveness but amplify noise
Longer periods reduce noise but increase lag

Implementation notes in tradedesk¶

EMA is tick-driven or candle-driven depending on usage
EMA requires warmup equal to its period length
EMA is commonly used as a building block for MACD

Simple Moving Average (SMA)¶

Conceptual overview¶

A Simple Moving Average (SMA) is the arithmetic mean of the last N prices.

It provides a baseline view of trend direction with maximal smoothing and lag.

Mathematical definition¶

SMA(t) = (P(t) + P(t−1) + … + P(t−N+1)) / N

Interpretation¶

SMA smooths price but reacts slowly to change
Crossovers are often used as trend filters

Implementation notes in tradedesk¶

SMA is candle-driven in most use cases
SMA must be fully warmed before use

Relative Strength Index (RSI)¶

Conceptual overview¶

RSI is a momentum oscillator that measures the speed and magnitude of recent price changes.

It is commonly used to identify overbought and oversold conditions.

Mathematical definition¶

RSI compares average gains to average losses over N periods and normalises the result to a 0–100 scale.

Interpretation¶

RSI > 70 → overbought
RSI < 30 → oversold

Thresholds are context-dependent and should not be treated as absolute.

Implementation notes in tradedesk¶

RSI is candle-driven
RSI requires sufficient warmup to stabilise averages

Average Directional Index (ADX)¶

Conceptual overview¶

ADX measures trend strength, not direction.

It is often used as a filter to determine whether trend-following signals are likely to be effective.

Mathematical definition¶

ADX is derived from the directional movement indicators (+DI and −DI), which are based on directional price changes.

Interpretation¶

Low ADX → weak or ranging market
High ADX → strong trend (direction determined elsewhere)

Implementation notes in tradedesk¶

ADX is candle-driven
ADX is often used as a gating condition rather than a signal

Bollinger Bands¶

Conceptual overview¶

Bollinger Bands describe price volatility relative to a moving average.

They consist of:

a middle moving average
an upper and lower band offset by standard deviation

Mathematical definition¶

Upper = MA + (k × σ)
Lower = MA − (k × σ)

where σ is the standard deviation over the lookback window.

Interpretation¶

Band expansion → increasing volatility
Band contraction → volatility compression

Implementation notes in tradedesk¶

Bollinger Bands are candle-driven
Interpretation depends heavily on parameter selection

Commodity Channel Index (CCI)¶

Conceptual overview¶

CCI measures deviation of price from its statistical mean.

It is often used to identify cyclical turning points.

Mathematical definition¶

CCI compares typical price to its moving average, scaled by mean deviation.

Interpretation¶

High positive CCI → price above historical norm
High negative CCI → price below historical norm

Implementation notes in tradedesk¶

CCI is candle-driven
Sensitive to outliers

Stochastic Oscillator¶

Conceptual overview¶

The stochastic oscillator compares the current close to the recent high–low range.

It assumes momentum precedes price.

Mathematical definition¶

%K = (Close − LowestLow) / (HighestHigh − LowestLow)

Interpretation¶

Values near extremes indicate potential reversals

Implementation notes in tradedesk¶

Candle-driven
Often smoothed with a moving average

On-Balance Volume (OBV)¶

Conceptual overview¶

OBV accumulates volume based on price direction.

It attempts to detect accumulation and distribution.

Mathematical definition¶

If price closes up → add volume
If price closes down → subtract volume

Interpretation¶

Rising OBV supports bullish price action
Divergence may indicate weakening trends

Implementation notes in tradedesk¶

Candle-driven
Volume quality is critical

Volume Weighted Average Price (VWAP)¶

Conceptual overview¶

VWAP represents the average price traded, weighted by volume.

It is commonly used as an institutional benchmark.

Mathematical definition¶

VWAP = Σ(price × volume) / Σ(volume)

Interpretation¶

Price above VWAP → bullish bias
Price below VWAP → bearish bias

Implementation notes in tradedesk¶

VWAP is session-dependent
Requires careful reset semantics

License¶

Licensed under the Apache License, Version 2.0. See: https://www.apache.org/licenses/LICENSE-2.0