Most trade tagging systems fail for the same reason. They collect labels instead of evidence. You add a tag for "breakout", another for "revenge trade", another for "London session", and six months later you have four hundred trades wearing forty different tags and no clear idea which combination is quietly costing you money. Tagging is not the point. Comparison is. A tag only earns its place if it can split your trades into groups you can measure against each other. If your tags cannot answer a question like "is my A-setup actually better than my B-setup, and by how much", they are decoration. This is how to build a system that answers that kind of question, and keeps answering it as your sample grows. A tag is a question you want to answer later Before you create any tag, finish this sentence: "I want to know whether ___ changes my results." Whether trading before the New York open helps or hurts. Whether moving my stop early is a habit or a rare slip. Whether my breakout entries beat my pullback entries once you account for how often each one fires. Every tag should map to one of those questions. If you cannot name the question a tag is meant to answer, do not add it. This single rule kills most tag bloat before it starts, because "it felt relevant" is not a question, it is a reflex. Start with fewer tags than you think you need New journalers tend to build a fifty-tag taxonomy on day one, then abandon it by week three because tagging every trade becomes a chore. The opposite approach works better. Start with two or three dimensions, tag consistently for fifty trades, and only add a tag when you notice a pattern you genuinely cannot describe with what you already have. Consistency beats granularity. A tag applied to 80% of your trades tells you far more than a tag you used enthusiastically for a fortnight and then forgot. The whole system rests on comparable groups, and you cannot compare groups that were tagged by different versions of yourself. Group your tags into dimensions The trick that makes tags readable is treating them as categories rather than a flat pile of keywords. Think in dimensions, where each dimension holds a small set of values, and most of the time you pick exactly one value per dimension per trade. Setup What pattern got you in. Breakout, pullback, range reversal, news reaction, whatever your playbook actually contains. Keep this list short and mutually exclusive. If a trade genuinely does not fit any named setup, that is useful information, so give it an "off-plan" value rather than inventing a new setup on the spot. Market context The conditions around the trade, not the trade itself. Trend versus range, session, volatility regime, whether a major data release was nearby. Context tags are where a lot of hidden edge lives, because a setup that prints money in a trending market can bleed slowly in chop, and a flat overall win rate hides that completely. This is also where fundamentals-aware traders separate a clean setup from one fighting the macro backdrop, which is worth reading the fundamentals for before you blame the pattern. Execution quality Did you follow the plan or not. This is the most valuable dimension and the one traders avoid, because it forces honesty. Tag whether the entry matched your rules, whether you sized correctly, and whether you managed the trade as planned or improvised. A profitable trade you took by breaking your own rules is not a win, it is a warning. Mistakes A separate multi-select dimension for specific errors. Chased the entry. Moved the stop. Took profit early. Added to a loser. These are the tags that pay for the whole exercise, because a mistake tag applied across enough trades turns a vague feeling of "I keep messing up my exits" into a number you can act on. Single-select where you can, multi-select where you must Setup, context, and execution work best as single-select. One setup, one session, one honest verdict on whether you followed the plan. Mistakes and secondary conditions can be multi-select, because a single trade can carry more than one leak at once (chased the entry and moved the stop). The reason matters. Single-select dimensions partition your trades cleanly, so every trade lands in exactly one bucket and the buckets add up to your whole account. Multi-select tags overlap, which is fine for spotting recurring errors but useless for clean comparisons. Mixing the two without noticing is why so many journals produce numbers that do not reconcile. Reading the tags without fooling yourself Once trades are tagged, the temptation is to sort by win rate and crown a winner. Do not. Win rate on its own is close to meaningless if the sample is small or the payoffs differ. A tag with a 40% win rate and an average winner three times the size of the average loser beats a 65% tag that scalps tiny gains and occasionally gives them all back. Judge each tag group on expectancy, not hit rate. Expectancy folds win rate and average result into one figure, which is the whole argument in R-multiples versus win rate , and it is the honest way to rank setups that behave differently. Alongside it, always look at the sample size. Ten trades under a tag can say almost anything. Fifty starts to mean something. If a tag has five trades, treat any conclusion as a guess wearing a lab coat. Two comparisons do most of the work: Setup against setup. Which of your named patterns actually carries the account, once expectancy and frequency are both in view. Same setup, different context. Take your best setup and split it by session or by trend versus range. This is where you usually find that the edge is not the setup at all, it is the setup in a specific condition . Common tagging mistakes A few patterns quietly wreck otherwise good systems. Outcome tags disguised as process tags. Tagging a trade "good" or "bad" after you know the result teaches you nothing except that winners feel good. Tag the process, not the profit. Tags nobody can define. If "clean setup" means something different every time you use it, it cannot group anything. Write a one-line definition for each tag and stick to it. Retro-fitting tags to explain a loss. Adding a fresh mistake tag only after a trade goes wrong biases the whole record. Tag at entry and at exit based on rules, not on how it felt afterwards. Too many setups. If you have fifteen setups, you probably have three setups and twelve variations you have not admitted are the same thing. Turn the tags into a decision A tagging system that never changes your behaviour is a diary, not an edge finder. Every few weeks, pull the tag breakdown and ask three questions. Which setup earns its risk. Which context turns a good setup bad. Which mistake tag shows up most often, and what would removing it do to your bottom line. Usually one or two answers are loud enough to act on immediately. Stop trading your worst-expectancy setup for a month and watch what happens to the account. Cut trades in the context where your best setup goes flat. Put a hard rule around your most frequent mistake. This is the practical core of learning to find your edge from your journal , and tags are simply the filing system that makes the edge visible. In TradeSave+ you attach these tags as you log each trade, then filter and compare expectancy across any tag or combination, so the setup-versus-context question becomes a two-click view rather than a spreadsheet afternoon. The software does the grouping. Your job is the honesty that makes the groups worth reading. Build the taxonomy small, tag it the same way every time, judge each group on expectancy with an eye on sample size, and act on the loudest signal. Do that for a few hundred trades and your tags stop being labels. They become the map of where your money actually comes from, and where it leaks out.
How to Build a Trade Tagging System That Surfaces Edges
Most tagging systems collect labels, not evidence. Here is how to tag trades so your real edges (and leaks) actually show up in the data.
Most trade tagging systems fail for the same reason. They collect labels instead of evidence. You add a tag for "breakout", another for "revenge trade", another for "London session", and six months later you have four hundred trades wearing forty different tags and no clear idea which combination is quietly costing you money. Tagging is not the point. Comparison is. A tag only earns its place if it can split your trades into groups you can measure against each other. If your tags cannot answer a question like "is my A-setup actually better than my B-setup, and by how much", they are decoration. This is how to build a system that answers that kind of question, and keeps answering it as your sample grows. A tag is a question you want to answer later Before you create any tag, finish this sentence: "I want to know whether ___ changes my results." Whether trading before the New York open helps or hurts. Whether moving my stop early is a habit or a rare slip. Whether my breakout entries beat my pullback entries once you account for how often each one fires. Every tag should map to one of those questions. If you cannot name the question a tag is meant to answer, do not add it. This single rule kills most tag bloat before it starts, because "it felt relevant" is not a question, it is a reflex. Start with fewer tags than you think you need New journalers tend to build a fifty-tag taxonomy on day one, then abandon it by week three because tagging every trade becomes a chore. The opposite approach works better. Start with two or three dimensions, tag consistently for fifty trades, and only add a tag when you notice a pattern you genuinely cannot describe with what you already have. Consistency beats granularity. A tag applied to 80% of your trades tells you far more than a tag you used enthusiastically for a fortnight and then forgot. The whole system rests on comparable groups, and you cannot compare groups that were tagged by different versions of yourself. Group your tags into dimensions The trick that makes tags readable is treating them as categories rather than a flat pile of keywords. Think in dimensions, where each dimension holds a small set of values, and most of the time you pick exactly one value per dimension per trade. Setup What pattern got you in. Breakout, pullback, range reversal, news reaction, whatever your playbook actually contains. Keep this list short and mutually exclusive. If a trade genuinely does not fit any named setup, that is useful information, so give it an "off-plan" value rather than inventing a new setup on the spot. Market context The conditions around the trade, not the trade itself. Trend versus range, session, volatility regime, whether a major data release was nearby. Context tags are where a lot of hidden edge lives, because a setup that prints money in a trending market can bleed slowly in chop, and a flat overall win rate hides that completely. This is also where fundamentals-aware traders separate a clean setup from one fighting the macro backdrop, which is worth reading the fundamentals for before you blame the pattern. Execution quality Did you follow the plan or not. This is the most valuable dimension and the one traders avoid, because it forces honesty. Tag whether the entry matched your rules, whether you sized correctly, and whether you managed the trade as planned or improvised. A profitable trade you took by breaking your own rules is not a win, it is a warning. Mistakes A separate multi-select dimension for specific errors. Chased the entry. Moved the stop. Took profit early. Added to a loser. These are the tags that pay for the whole exercise, because a mistake tag applied across enough trades turns a vague feeling of "I keep messing up my exits" into a number you can act on. Single-select where you can, multi-select where you must Setup, context, and execution work best as single-select. One setup, one session, one honest verdict on whether you followed the plan. Mistakes and secondary conditions can be multi-select, because a single trade can carry more than one leak at once (chased the entry and moved the stop). The reason matters. Single-select dimensions partition your trades cleanly, so every trade lands in exactly one bucket and the buckets add up to your whole account. Multi-select tags overlap, which is fine for spotting recurring errors but useless for clean comparisons. Mixing the two without noticing is why so many journals produce numbers that do not reconcile. Reading the tags without fooling yourself Once trades are tagged, the temptation is to sort by win rate and crown a winner. Do not. Win rate on its own is close to meaningless if the sample is small or the payoffs differ. A tag with a 40% win rate and an average winner three times the size of the average loser beats a 65% tag that scalps tiny gains and occasionally gives them all back. Judge each tag group on expectancy, not hit rate. Expectancy folds win rate and average result into one figure, which is the whole argument in R-multiples versus win rate , and it is the honest way to rank setups that behave differently. Alongside it, always look at the sample size. Ten trades under a tag can say almost anything. Fifty starts to mean something. If a tag has five trades, treat any conclusion as a guess wearing a lab coat. Two comparisons do most of the work: Setup against setup. Which of your named patterns actually carries the account, once expectancy and frequency are both in view. Same setup, different context. Take your best setup and split it by session or by trend versus range. This is where you usually find that the edge is not the setup at all, it is the setup in a specific condition . Common tagging mistakes A few patterns quietly wreck otherwise good systems. Outcome tags disguised as process tags. Tagging a trade "good" or "bad" after you know the result teaches you nothing except that winners feel good. Tag the process, not the profit. Tags nobody can define. If "clean setup" means something different every time you use it, it cannot group anything. Write a one-line definition for each tag and stick to it. Retro-fitting tags to explain a loss. Adding a fresh mistake tag only after a trade goes wrong biases the whole record. Tag at entry and at exit based on rules, not on how it felt afterwards. Too many setups. If you have fifteen setups, you probably have three setups and twelve variations you have not admitted are the same thing. Turn the tags into a decision A tagging system that never changes your behaviour is a diary, not an edge finder. Every few weeks, pull the tag breakdown and ask three questions. Which setup earns its risk. Which context turns a good setup bad. Which mistake tag shows up most often, and what would removing it do to your bottom line. Usually one or two answers are loud enough to act on immediately. Stop trading your worst-expectancy setup for a month and watch what happens to the account. Cut trades in the context where your best setup goes flat. Put a hard rule around your most frequent mistake. This is the practical core of learning to find your edge from your journal , and tags are simply the filing system that makes the edge visible. In TradeSave+ you attach these tags as you log each trade, then filter and compare expectancy across any tag or combination, so the setup-versus-context question becomes a two-click view rather than a spreadsheet afternoon. The software does the grouping. Your job is the honesty that makes the groups worth reading. Build the taxonomy small, tag it the same way every time, judge each group on expectancy with an eye on sample size, and act on the loudest signal. Do that for a few hundred trades and your tags stop being labels. They become the map of where your money actually comes from, and where it leaks out.