Best Practices for Using NiveshBuddy
NiveshBuddy gives you evidence about a strategy; your habits decide whether you read it honestly. The do's, the don'ts, and a short checklist for building, backtesting and simulating without fooling yourself.
What You’ll Learn
By the end of this article, you’ll understand:
- Why a repeatable process matters more than any one result
- The habits that make backtests and experiments trustworthy
- How to run a simulation without undermining it
- A checklist to run before acting on a strategy
Why it matters
NiveshBuddy gives you evidence: how a set of rules behaved on past data in a backtest, how nearby versions of it behaved in an Experiment, and how it behaves going forward in a Simulation — a paper portfolio run with virtual money. None of that evidence reads itself. The same backtest can inform a careful decision or justify a careless one, and the difference lies almost entirely in the habits of the person reading it.
That is why a repeatable process matters more than any single result. One strong backtest can be luck, a favourable window, or the product of tweaking until the numbers looked right. A process you follow the same way every time — and can explain afterwards — is what lets you tell those apart.
The loop this article is built around is simple:
One responsible loop
- BuildA Quick Strategy or a Custom Strategy, with rules you can state
- BacktestHow the rules behaved on historical data
- CompareCompare Results, on a like-for-like basis
- Explore (optional)An Experiment, on plans that include it
- SimulateVirtual money, forward in time
- Decide independentlyAt your own broker, on your own judgement
The habits below are grouped by where in that loop they apply. The general mistakes beginners make — reacting to headlines, overtrading, restarting after a bad month — are covered in Common Investing Mistakes Every Beginner Should Avoid. This piece is about using the product well.
Do's
Build
Most of the damage done to a strategy happens before it is ever tested, by making it too complicated to understand. Start where the product makes it easy.
Start simple
A Quick Strategy is a working template: three choices, a full set of preset rules, and no Strategy credits spent. You can read exactly what it set before you change anything.
Change one rule at a time
Customize opens the strategy in the Custom Strategy Builder with its current values filled in. Change one setting, re-run, and you know what caused any difference. Each re-run spends one Strategy credit.
Save As to keep both
Save As copies a strategy under a new name and spends no credit. Copy first if you want the original and the changed version side by side.
The reason for one change at a time is attribution. Change the universe, the holding count and the rebalance frequency together, and a better result tells you only that something helped. Change one, and the result is about that one thing. Compare Results makes this easy to check: its Settings section opens on differences only, so a fair comparison of two versions should show a single row.
It also helps to be able to say what the strategy does in a sentence before you test it — "the fifteen strongest Nifty 200 stocks by six-month momentum, rebalanced monthly", for instance. If you can't, the rules are probably doing things you haven't decided on.
Backtest
A backtest describes one historical result. These habits are about not asking it to say more than that.
Read risk with return
Look at Max Drawdown and Longest Drawdown Days next to CAGR — how far it fell, and how long it stayed below a previous high. The Score and Key takeaway are generated summaries of one result, not a judgement on whether it suits you.
Compare like with like
Put two results side by side in Compare Results only when they share a test period and a Transaction Cost. The Comparison Quality banner warns when periods differ, and the Settings section shows any other difference.
Test on another period
Before trusting a change, run it over a Custom Backtest Period it was not found on. An improvement that holds on one stretch of history and vanishes on the next belonged to the stretch.
Treat first place as a lead
An Experiment always ranks something first. Look at whether the close variants around it agree, then test it on a period it was not ranked on before you rely on it.
Set a cost you would pay
Transaction Cost is your own assumption, charged on every buy and every sell, and it starts at zero. Quick Strategy backtests run at zero, so Customize to set one that reflects what you expect to pay.
Here is how those habits fit together, in a hypothetical, illustrative case. Suppose a monthly strategy's backtest shows more trades than you expected, and you suspect it sells holdings too early. You use Save As to keep the original, widen the copy's Exit Rank — nothing else — and re-run it over the same five years at the same Transaction Cost. The copy trades less and shows a slightly higher CAGR. Compare Results shows one row in Settings and no Comparison Quality warning, so the comparison is fair.
That is where most people stop, and it is too early. Re-run both versions on the first half of the window, then on the second. If the lower trade count holds in both halves but the higher return appears in only one, you have learned something useful: the change reliably reduces trading, and the return gain probably belonged to that stretch of history. You keep the change for the reason you made it, and stop expecting more from it than that.
An Experiment does the same thing at scale, which is why it needs the same discipline. Ranking many configurations on one window will always put one of them first, partly because it suited that window. The close variants around it — and a period it was not ranked on — tell you how much of its lead to believe.
Simulate
A simulation is only useful as evidence if it is left to run the strategy. Each of these habits protects that.
Give it several rebalances
The first batch of scheduled trades is all buys, because the portfolio starts in cash; only later batches show what the rules change. The simulation's own Key takeaway reads Too early to judge for its first 30 days.
Let scheduled trades run
Scheduled Trades run automatically at the rebalance day's close unless you reject them. Reject sparingly, and keep your own note of why each time.
Record what you actually did
If you act at your broker, record it with Add, Edit, Sell or Replace Position. These record your trades without changing the strategy, so the simulation stays aligned with reality.
Review on a cadence
Check Scheduled Trades before each rebalance, in the simulation or the Action Center. Between rebalances there is nothing to decide, and daily checking invites changes you did not plan.
Write down every change
Edit Settings applies from the next rebalance and is logged in Configuration History. The log records what changed; add your own note of why, while you still remember.
A note on timing, because it shapes what reviewing on a cadence means. Signals for a rebalance come from the previous trading day's closing data, and the trades run at the rebalance day's official close. Your window to reject is between those two points — there is nothing intraday to watch.
Keep two kinds of activity separate in your head, because the simulation keeps them separate in its record. A strategy-generated trade is what the rules asked for. A portfolio adjustment is something you recorded — a position you added carries a Manual pill, for example. Both are legitimate. But a simulation full of rejects and manual changes is no longer showing you the strategy; it is showing you your own decisions, and its results say more about those than about the rules.
Don'ts
Each of these feels like diligence at the time. Each one quietly makes the evidence less trustworthy.
| Don't | Why | Instead |
|---|---|---|
| Chase the highest CAGR | Return alone ignores how far and how long the strategy fell to earn it | Read CAGR alongside Max Drawdown and Longest Drawdown Days, and ask whether you could have sat through them |
| Stack rules until it fits | Every rule added to rescue a backtest fits that history a little more and the future no better | Add a rule only when you can say, before testing, what it should do |
| Read one window as proof | One period is one market regime, and the result may belong to it | Re-run on a different Custom Backtest Period and see what survives |
| Reject scheduled trades by reflex | Each reject replaces the strategy's rule with a judgement made in the moment, and the record stops showing the strategy | Reject rarely, with a reason you could defend later |
| Judge a simulation on its first batch | The first batch is all buys, building the portfolio from cash; it says nothing yet about how the rules perform | Wait several rebalances before drawing any conclusion |
| Treat a ranking, a Key takeaway or an AI assistant's answer as advice | Latest Strategy Rankings are your strategy's output, a Key takeaway describes one result, and an assistant's answer is generated by that assistant | Use them to understand what happened, then decide for yourself |
The last row deserves emphasis. The product labels its rankings "for transparency, not recommendations", and if you connect an AI assistant, what it says comes from that assistant, not from NiveshBuddy. None of these tells you what to buy or sell.
It is also worth keeping a backtest and a simulation distinct when you weigh them up. A backtest replays the rules over history that has already happened, and you can re-run it as often as you like. A simulation moves forward one rebalance at a time through markets nobody has seen yet — it starts today or on the next trading day, never in the past. Agreement between the two is encouraging; disagreement is information, not a fault in either. Neither predicts what the strategy will do next.
Key takeaways
A process outlasts a result. A single strong backtest can be luck or a favourable window. A loop you follow the same way every time is what lets you tell the difference.
Change one thing, then compare fairly. One rule at a time, on the same period and the same cost — otherwise you cannot say what caused a difference.
A result has to survive a period it was not found on. That goes for a manual change and for an Experiment's first place alike.
Let the simulation run the strategy. Scheduled trades run unless you reject them; reject rarely, record what you really did, and judge it only after several rebalances.
Before you act
Backtests and simulations can help you evaluate a strategy, but they do not predict how it will behave from here. Before you act on one at your own broker, run through these questions.
- Can I state the strategy's rule in one sentence?
- Did I compare results on the same period and the same Transaction Cost?
- Has the change held on a period it wasn't found on?
- Has the simulation run for several rebalances?
- Do I know why I rejected each scheduled trade I rejected?
- Would I be comfortable living through the worst drawdown with real money?
- Is this decision mine, made independently?
A "no" to any of these is not a reason to give up on a strategy. It is a sign that the evidence is not finished yet — and that the next step is in NiveshBuddy, not at your broker.