How the challenge works

Your journey at a glance

All times are India Standard Time (IST). UAE students: subtract 1 hour 30 minutes for Gulf Standard Time (GST).

The elimination round, minute by minute

Sunday 22 November 2026, 11:00 AM–1:00 PM IST (9:30–11:30 AM GST). Online, on CMT Association's challenge platform. 120 minutes.

Log in a few minutes before 11:00 AM IST.

The round runs on one clock for everyone. Part 1 locks at 11:40 AM IST. Any rulebook you have not sealed seals automatically at 12:40 PM IST. Everything locks at 1:00 PM IST. You may finish a part early: submitting Part 1 opens Part 2 at once, and sealing your fund opens Part 3 at once. Time lost to a late start or a dropped connection is not added back, except under the outage rule (see Rules of fair play).

Briefing (10 minutes)

What you see: your client's mandate and the universe of investments you can choose from.

The client mandate:

"A family is saving for their daughter's university fees in ten years. Grow the money, but they cannot accept a fall of more than a quarter of its value."

The universe: ten US industry groups, plus cash.

Your fund holds some of these industries and keeps the rest in cash. Cash earns the US Treasury bill rate. The whole US market serves as the benchmark.

What you do: read the mandate twice. Write down the one number that matters most: a 25% maximum fall.

Scored: nothing. This is reading time.

Part 1: Spot the AI's mistake (30 minutes)

What you see: 10 charts, each with an AI's reading of the chart. The charts are constructed teaching charts with price and volume, not the industry data you use in Part 2. Five of your charts contain a mistake and five are correct. They appear in random order, and you are not told which are which.

What you do: for each chart, choose one:

• Agree if the AI's reading is correct.
• Flag if the AI made a mistake. Then choose the type of mistake from this list:

You may also add a one-line reason. It is not scored.

How long: 30 minutes. Part 1 locks at 11:40 AM IST. That is about 3 minutes per chart.

Scored: 3 points per chart, 30 points in total. See Section 7.

Common mistakes:

• Flagging everything. Half the charts are correct. A wrong flag on a correct chart scores 0.
• Trusting the AI because it sounds confident. Look at the chart itself.
• Spending 10 minutes on one chart. Move on and come back if time allows.
• Ignoring the axis labels. Check the scale before you judge the move.

Part 2: Build your fund (60 minutes)

What you see: the rulebook builder, with five building blocks. Each block starts at its default. You also see test results for the practice decade, 2010–2019.

What you do: choose a value for each block, run a test, study the result, and adjust. You have unlimited test runs. When you are satisfied, seal your fund.

The five building blocks

What each block does

1. Selection. On each rebalance week, the fund ranks all ten industries by their return over the last N weeks and picks the top K. The hurdle adds a test. With "Must beat the market," a pick that did worse than the market goes to cash instead. With "Must have a positive return," a pick that lost value over N weeks goes to cash instead.
2. Trend filter. Every week, the fund checks each industry it holds. If an industry falls below its X-week moving average, the fund moves that industry's share to cash. It stays in cash until the next rebalance, even if the industry recovers. A newly picked industry that is already below its average goes straight to cash.
3. Market regime. Every week, the fund checks the whole market. If the market falls below its M-week moving average, the fund cuts all industry holdings to Y% of their normal size and keeps the rest in cash. With Y at 0%, the fund moves fully to cash. Full holdings return when the market closes back above its average.
4. Position sizing. With equal weight, each chosen industry gets the same share. With inverse volatility, industries that moved more calmly over the last 26 weeks get a larger share. The platform measures this as the sample standard deviation of the last 26 weekly returns, with a floor of 0.1% a week.
5. Rebalancing. The fund reviews its picks every R weeks.

How the fund operates

• The fund starts in cash.
• It looks at data up to the end of a week, makes its changes at the end of the next week, and earns returns on its new holdings from the week after that. The fund cannot act on information it does not yet have.
• Every change in holdings costs 0.10% of the amount moved.
• Cash earns the weekly US Treasury bill return.

The plain-English summary before you seal

Before you seal, the platform shows your rulebook as a short plain-English summary. For example: "Every 8 weeks, hold the 3 industries with the best 26-week return, in equal shares. Move any industry to cash if it falls below its 30-week average." Read it carefully. If it does not say what you meant, change your blocks before you seal. In the final round, you must be able to explain every rule in it.

Sealing and the hidden windows

When you seal, your fund is final. You cannot change it.

The practice decade, 2010–2019, is only for testing. Your sealed fund is scored on three hidden five-year windows. You have never been told which periods they are, and you will not be told during the elimination round. The hidden windows test whether your rules work in market conditions you did not tune them for. This is why overfitting hurts and sensible rules help.

How long: 60 minutes.

Scored: 30 points, based on how your sealed fund performs in the hidden windows. See Section 7.

Common mistakes:

• Chasing the highest growth on 2010–2019. The hidden windows will differ.
• Ignoring the mandate. Check your fund's maximum fall on every test.
• Changing five things at once. Change one block, test, and learn what it does.
• Forgetting to seal. An unsealed rulebook seals automatically at 12:40 PM IST, exactly as it stands.
• Waiting until the last minute. If two students tie on every other count, the earlier seal time wins.

Part 3: Explain your fund (15 minutes)

What you see: four structured questions about your sealed fund, plus space for an optional note of up to 60 words.

What you do:

How long: 15 minutes.

Scored: 40 points. The optional 60-word note is not scored. Jurors see it for the top 10.

Common mistakes:

• Answering Q2 in general terms. The question is about your fund and your rules. The platform knows how your fund behaves in each scenario.
• Accepting AI advice because it sounds reasonable. Ask whether it helps the client within the mandate.
• Guessing on Q4. Look at your fund's maximum drawdown for 2010–2019 and compare it with 25%.

Buffer (5 minutes)

Use this time to check your answers. Everything locks at 1:00 PM IST. Any rulebook not sealed by 12:40 PM IST has already sealed automatically.

How the elimination round is scored

Scoring is fully automatic. The total is 100 points.

Part 1, chart by chart

What "median" and "ranked" mean

Your fund runs through three hidden windows, so it gets three results for each measure. The median is the middle one. If your excess return per drawdown is 0.2, 0.5 and 0.9, your median is 0.5. The median ignores one lucky or unlucky window and rewards a fund that does well consistently.

Your median is then ranked against every other student's median. Rank 1 is best. Points = maximum × (n − rank) ÷ (n − 1), where n is the number of sealed funds. The best fund scores the full maximum and the last scores 0. Tied students share the average rank. Points are rounded to two decimals. For example, with 101 sealed funds, rank 26 on a 12-point measure scores 12 × 75 ÷ 100 = 9.00. You compete on how your fund compares with others in the same conditions, not on a fixed target.

Why judgment counts more than raw return

Raw return earns no points on its own. A fund that doubles in value but falls 50% along the way fails this client. The scoring rewards three things a real fund manager needs:

• Growth for the risk taken (excess return per drawdown).
• Protection when markets fall (relative drawdown and the mandate).
• Sound judgment about charts, rules and AI advice (Parts 1 and 3, 70 points together).

A student who builds a steady fund, keeps the mandate and spots the AI's mistakes beats a student who simply chases the highest number.

Tie-breaks

If two students have the same total, the tie breaks in this order:

1. Higher Part 1 score.
2. Higher Part 3 score.
3. Earlier seal time.

The top 30 students reach the final. The shortlist is published by 6:00 PM IST on 22 November 2026. Appeals against elimination close at 6:00 PM IST on 23 November 2026 (24 hours). Finalists are confirmed by 8:00 PM IST that day.

The final round for the top 30

Sunday 29 November 2026, start time set by Edept. Online. About 2 hours 10 minutes. Results the same day.

The reveal

Each finalist sees how their sealed fund performed in the three hidden windows. The host names the periods.

The stress room

All 30 finalists face the same 26-week market crisis. Its dates are hidden, and the data starts at 100. You move through it in four checkpoints of 10 minutes each.

At each checkpoint you see:

• Charts up to that week.
• The holdings of a model fund, the Class Fund. Its rules: hold the top 3 industries by 26-week return, use a 30-week trend filter, equal weight, rebalance every 4 weeks.
• A memo from an AI analyst.

At each checkpoint you decide three things:

1. One action: follow the rules · cut exposure to half · sell one named industry.
2. The memo: accept it or reject it. Some memos are wrong. Nobody is told how many.
3. Your reason: pick the best reason for your decision from four choices.

The defenses

The top 10 finalists each defend their fund before the two judges:

• Joel Pannikot, CMT Association
• Brinda Kotak, Apexr

Each defense lasts 5 minutes: a 2-minute pitch, then 3 minutes of questions. Explain what your rules do, why you chose them, and how your fund protects the client. You must be able to explain every rule in your fund.

Final scoring (100 points)

• The top 10 for the defenses are chosen on elimination score (20) plus stress room (40), out of 60.
• Places 1–10 are ranked on the full 100 points.
• Places 11–30 are ranked on elimination score plus stress room.
• Ties for places 1–30 break on stress room score, then defense understanding (places 1–10 only), then elimination score, then elimination Part 1.
• A tie at tenth place for the defenses breaks on stress room score, then elimination score.
• Appeals against the final results close 48 hours after the results.

Winners

The top 20 finalists win in the Finance track.

Edept sets and pays all prizes and scholarships. Edept runs the recognition event at the WUD campus, Sonipat, on 5 December 2026 (tentative).