Analysis
Notational analysis in sport: methods, templates and pitfalls
· 10 min read
Notational analysis is the systematic recording of discrete match events — passes, shots, tackles, serves, turnovers — so that performance can be described in counts, frequencies and sequences rather than impressions. It is the oldest branch of performance analysis and still the most accessible: a pen, a sheet of paper and a defined event list are enough to start.
It is also the branch most often done badly, in one specific way that this article is mostly about.
What notational analysis actually captures
Four dimensions. Most beginners record two of them.
Frequency. How often it happened. Fourteen turnovers.
Outcome. Did it work. Nine of fourteen recovered within five seconds.
Those two are easy and they are where most coaches stop. The insight lives in the other two:
Location. Where on the pitch or court. Eleven of the fourteen turnovers in the left channel. That is a coachable finding; "fourteen turnovers" is not.
Sequence. What came before and after. Nine of the eleven left-channel turnovers followed a switch of play from the right. Now you know what to train, and it is not the left channel.
Adding location and sequence roughly doubles the coding effort and multiplies what the data can tell you. If you record only frequency and outcome, you will produce numbers that describe the match accurately and suggest nothing.
Hand notation vs computerised notation
| Hand notation | Computerised / video-linked | |
|---|---|---|
| Setup cost | A sheet of paper | Software licence, learning time |
| Speed live | Fast once practised | Slower live, faster in review |
| Re-checking | Hard: you cannot revisit the moment | Every data point is a clip |
| Reliability testing | Possible but laborious | Straightforward |
| Showing a player | Nothing to show but a number | The clip is the evidence |
| Sharing | Photograph of a sheet | Export, clip links |
The honest summary: hand notation is genuinely sufficient for finding patterns, and genuinely useless for changing behaviour. A tally sheet can tell you the left channel is a problem. It cannot show a fifteen-year-old the four occasions when it was them, and that is what moves a player.
That is why the natural progression is hand notation to establish that something is worth working on, then video-linked coding once you know what you are looking for.
What Sidetalk does and does not do here
To be direct, because this article will be read by people evaluating tools: Sidetalk is not a notational analysis system. It does not count events, code them, tally them or export counts. There is no event taxonomy to configure.
The markers it does have are structural — kick-off, half time, goals, and a generic marker for training — and they exist to establish match phase and to anchor the video clock. They are not an event-coding scheme, and using them as one would give you a very short list of very blunt data.
What Sidetalk records is the layer that notation cannot capture: your own commentary while it happened, attributed to individual players, and timestamped against your footage. It answers "what did I say about Lucas in the 23rd minute", not "how many turnovers did we have". Those are complementary jobs, and if what you need is counts, use a notation tool. The templates below are free, and so are several good apps.
Building your coding system
Five steps, in this order.
1. Start from the coaching question, not the event list. "Where do we lose the ball in build-up?" comes first. The events you code are whatever answers it. Coaches who start by listing events end up with a beautiful taxonomy and no findings.
2. Define each event operationally, in writing. This is the step that gets skipped and it is the one that matters. "Pressing action" means nothing. "A defender moves more than three metres towards an opponent in possession within two seconds of the ball arriving" means something, and, crucially, means the same thing next week.
3. Keep it to five to eight event types. Twelve is sustainable for about three weeks. This is close to a law. An abandoned coding system produces no baseline, and with no baseline the data you did collect is unusable.
4. Decide your outcome scale before you start. Binary success/fail beats a 1–5 scale you will apply inconsistently. If you cannot define the difference between a 3 and a 4 in one sentence, you do not have a scale, you have a mood.
5. Pilot on fifteen minutes of footage, then revise. You will discover that two of your definitions overlap and one is unobservable. Better now than after four matches of unusable data.
Reliability: the step that separates analysis from opinion
Almost no coaching blog covers this, and it is the difference between data and decoration.
Here is the failure mode. You code four matches. The number improves. You conclude the training worked. But what actually changed was your own coding. You got better at spotting the event, or looser about what counted, or you were tired in match one and sharp in match four. Your improvement is measurement drift, and it is invisible unless you test for it.
Two tests.
Intra-observer reliability. Code the same fifteen minutes of footage twice, at least a week apart, without looking at your first attempt. Compare the two.
Inter-observer reliability. Have someone else code the same passage against your written definitions. This is the harsher test and the more useful one.
The conventional thresholds in the published literature are agreement of at least 80%, or a kappa coefficient above 0.80. On the Landis and Koch benchmarks widely used in this field, 0.81 to 1.00 is treated as almost perfect agreement, and Robinson and O'Donoghue's weighted kappa method for performance analysis uses the same cut. Cooper and colleagues set out a simple procedure for testing data entered into sport performance analysis systems if you want the formal version.
You do not need to compute kappa to benefit from this. Counting how many of twenty events you and your assistant both coded the same way is enough to find out whether your definitions are real.
And when agreement is poor, the definitions are at fault, not the coder. That is the useful reframe. Poor agreement is a message that your written definition is ambiguous. Rewrite it, do not retrain the person.
A worked example
Illustrative figures, not customer data.
Question: where do we lose the ball building from the back?
Codes: build-up started (goalkeeper), build-up started (centre-back), turnover in own third, turnover in middle third, reached final third. Five codes, each with a written definition, zone recorded 1–6, pressed yes/no.
One match, tallied:
| Count | |
|---|---|
| Build-ups started | 31 |
| Turnover in own third | 7 |
| Turnover in middle third | 11 |
| Reached final third | 13 |
| Exit rate | 42% |
Frequency and outcome alone say: we lose it a lot in the middle third. Not actionable.
Adding location: nine of the eleven middle-third turnovers were in zone 4, the left side.
Adding sequence: seven of those nine came within two passes of the ball being switched from the right, and in six of the seven, the left-back was still high from the previous phase.
The finding is not "we are careless in midfield." It is: when we switch play left, our left-back has not recovered and the receiving midfielder has no outlet. That is one thing to train on Tuesday, and the re-measure next match is the same five codes.
That progression, from a useless true statement to a specific one, is what the location and sequence dimensions buy.
Free templates
Three sheets, no email required. Open them in Excel, Numbers or Sheets.
- General event tally sheet — sport-agnostic, with the definition column deliberately placed first
- Football build-up and turnover sheet — the worked example above, ready to fill in
- Court-sport rally outcome sheet — volleyball, basketball, handball, badminton
The definition column is first on purpose. It is the column everyone wants to skip and the one that determines whether the sheet produces anything.
Related analysis methods
Short definitions, because these are the terms that come up alongside notation and get muddled with it.
Time-motion analysis. Movement demands over time: distance, sprint counts, work-to-rest ratio. Answers "how hard is this position", not "what did they do". GPS or tracking video.
Needs analysis. Establishing the physical, technical and tactical demands of a sport or position before programming training. Done once per season or per position, not per match.
SWOT analysis in sport. Strategic and qualitative, usually about a team or opponent as a whole. Not event-level and not really analysis in the measurement sense.
Biomechanical or movement analysis. Technique at the level of the body: joint angles, contact times, frame-by-frame. Kinovea and Dartfish territory.
All four sit inside performance analysis, which is the umbrella term for the discipline.
Frequently asked questions
What is notational analysis in sport?
The systematic recording of discrete match events as countable data — frequency, location, sequence and outcome — so performance can be described in numbers rather than impressions. It is the oldest and most accessible branch of performance analysis.
What's the difference between notational analysis and performance analysis?
Performance analysis is the whole discipline, including biomechanics, time-motion work and the feedback loop. Notational analysis is the specific method of counting and coding events within it.
Do you need software for notational analysis?
No. Paper and a defined event list is a complete system, and the templates above are enough to start. Software becomes worthwhile when you want each data point to be a clip you can show a player, which paper cannot do.
What is time-motion analysis in sport?
Quantifying movement demand over a match or session — total distance, number of sprints, work-to-rest ratio — usually with GPS units or tracking video. It describes physical load rather than technical or tactical events.
How many events should I code?
Five to eight types. Fewer than five and you probably cannot answer your question; more than eight and the system tends not to survive a month, which costs you the baseline that makes any of it useful.
Where to go next
Once the tally sheet becomes the bottleneck — you know what the pattern is and you cannot show it to anyone — video-linked work is the next step: how to analyse sports video, step by step. For reading the numbers you have collected, how to interpret sports statistics.
Sources: Landis, J. R. and Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33(1), 159–174. · Robinson, G. and O'Donoghue, P. (2007). A weighted kappa statistic for reliability testing in performance analysis of sport. International Journal of Performance Analysis in Sport, 7(1), 12–19. · Cooper, S-M., Hughes, M., O'Donoghue, P. and Nevill, A. M. (2007). A simple statistical method for assessing the reliability of data entered into sport performance analysis systems. International Journal of Performance Analysis in Sport, 7(1), 87–109.