Understanding what is a trading signal
A trading signal, sometimes referred to as a trade signal, is an indicator or trigger to buy, sell, or hold a financial instrument.
The signal is a numerical value that falls into a predefined range with an upper and a lower limit. All buy/sell signals on SYGNAL use a standardized format ranging from -1.00 (strong sell) to +1.00 (strong buy) to ensure cross-modal comparison, easy strategy implementation, coherent decision-making, and simplified integration.
SYGNAL specializes in collecting and aggregating quantitative financial signals: the numerical output of complex mathematical and statistical models applied to financial instruments, representing past and present behavior and attempt to predict future behavior.
We select signals based on past alpha generation, from professional "Quants": quantitative managers, hedge funds, and data scientists.
What's in a trading signal
A trading signal is a standardized value expressing how bullish or bearish a quantitative model is about a given financial instrument. It is the output value of a signal model (quant model): the application of mathematical and statistical analysis based on quantitative theory to vast input-data quantities.
Input-data can range from historical prices to social media posts, satellite imagery, patent applications, and any other data-type used to model the asset's past and current behavior.
Signal models may differ in input-data and analytical methodology (mean-reversion, trend-following, global macro, etc.); however, the goal remains to express the output in a standardized and straightforward format: the signal.
Note: One should regard quantitative signals as an opinion generated by a quantitative model. The trade signal provides helpful information for evaluating a stock's or another financial instrument's attractiveness, but it is only one piece of information. One should always consider additional information before making a sound investment decision.
The signal scale
The maximum and minimum range of a signal is called the signal scale. SYGNAL standardizes all signals according to one of three scales:
|Scale||Max value||Min value||Used for|
|Long-Short||+1.00||-1.00||Trading: Buy, sell or hold an underlying instrument|
|Long||+1.00||0.00||Increasing exposure, hedge (for example, oil prices for an airline), or buy an underlying instrument|
|Short||0.00||-1.00||Reducing exposure, hedge, or sell (short) an underlying instrument|
The signal will always fall within the range of one of the scales mentioned above. The most common signal scale is long-short, with the vast majority of signals calculated according to this range. For more information about Short and Long-only scales, please refer to this article.
Interpreting a trade signal
The trading signal represents how bearish or bullish the quant model is about a given instrument. Generally speaking, positive signals express a bullish or optimistic opinion about an asset, while negative signals suggest a bearish or pessimistic outlook. Signals of 0.00 are neutral, neither bullish nor bearish:
As a general rule of thumb: positive signals mean that the model predicts that the instrument's value will increase in the future and that the trader should BUY. Negative signals indicate that the model is pessimistic about the instrument's future value and that the trader should SELL. Neutral signals suggest that the trader should do nothing.
Binary versus continuous signals
Signals can be binary or continuous:
|Binary signals||Signals containing only the maximum and minimum values on a signal scale: -1.00, 0.00, or +1.00||-1.00 = 100% negative conviction
0.00 = Neutral
1.00 = 100% positive conviction
|Continuous signals||Signals falling anywhere within the full range of the signal scale, indicating the degree to which the model is convinced about its prediction||-0.55 = 55% bearish
0.00 = Neutral
0.40 = 40% bullish
Binary signal models generate values of 0.00 and 1.00, or -1.00, with no intermediate values. Binary signals are not naturally inferior to non-binary trading signals. Instead, it is the signal manager's choice of how the output of their signal model should "look." Mostly it is the result of an in-depth analysis of the market structure that warrants the option for the right type of signal output.
Continuous signal models generate any value between -1.00 and 1.00 (for example -0.33 or +0.87). Therefore, the output value represents the degree to which the model is bullish or bearish about a particular instrument. For example, a signal of -0.10 suggests that the model is slightly bullish on the particular asset; likewise, a signal of 0.30 suggests that the model is relatively bullish about the asset.
Which are better, binary or continuous signals?
Both binary and continuous signal models have benefits and drawbacks. A binary signal typically has a high conviction for a trade, whereas the continuous signal is more nuanced in its expression. After their analysis, the signal manager ultimately decides which signal type best reflects the model.
The run frequency is the interval to run the quant model to incorporate new market information (i.e., price changes) and generate a new signal. Generally speaking, the run frequency is more frequent with shorter-term models and less frequent with longer-term models.
The following table defines the most common run frequencies:
|Intraday||Yes||Multiple times per day|
|Daily||Yes||One time per day, generally after market closing|
|Weekly||Yes||One time per week, generally after market closing|
|Monthly||Yes||One time per month|
|Quarterly||Yes||One time per quarter|
|Ad Hoc||No||Signal updated only when an input variable has changed (i.e., a company applies for a new patent, or the price of the asset hits a predefined point)|
A signal model using price as an input factor will usually take the price from the same point each day or hour. For most instruments, the price could either be the daily closing or opening price or the price at a specific time (on the hour or midnight UTC) each day, week, or month.
For signals running on an ad-hoc basis, the model generally triggers when the price hits a predetermined point or when another external event occurs (i.e., the company registers a new patent). Such models may run continuously, watching for market events or other patterns that may trigger a new signal generation.
The value, together with the expected accuracy, of a signal decays over time. As a general rule of thumb, a signal is most accurate when generated. As time passes from the generation point, new information such as price fluctuations, external events, or market factors will influence the signal's accuracy. Therefore, traders want to apply signals as close to the generation point as possible, assuming enough liquidity in the market.
How quickly a signal decays depends in part upon the signal's time horizon. Shorter-term signals decay faster than mid-to-long-term signals. It would not make sense to apply an intraday signal with 24 hours delay, but it may make sense to implement a mid-term signal over-the-day to limit slippage, as the signal value does not decrease as fast as an intraday signal.