Common Lumibot Strategy Errors in Data, Options, and Order Handling

# common mistakes


Common Mistakes and How to Avoid Them
======================================

.. meta::
   :description: This page documents the most common mistakes made when writing Lumibot strategies, along with the correct patterns to use instead.

This page documents the most common mistakes made when writing Lumibot strategies, along with the correct patterns to use instead.

Choosing an Intraday Order Duration
--------------------------------------------------------------------------------

``Strategy.create_order()`` defaults to ``time_in_force="gtc"``, while constructing
``Order`` directly defaults to ``"day"``. Specify the intended duration explicitly.
For example, an intraday market short can use:

.. code-block:: python

   hedge = self.create_order(
       Asset("SPY"), 100, Order.OrderSide.SELL_SHORT,
       order_type=Order.OrderType.MARKET, time_in_force="day",
   )
   self.submit_order(hedge)

Schwab may reject GTC short-sale orders for hard-to-borrow securities. DAY does
not guarantee acceptance or a fill; other borrowing, account and market constraints
still apply. Inspect the broker's rejection reason before retrying. Preserve
intentional GTC orders instead of changing every order's duration globally.

Restoring Asset Variables After a Restart
--------------------------------------------------------------------------------

Keep instruments as ``Asset`` objects in ``self.vars``. New scheduled-file and
database backups preserve their type, including assets nested in lists,
dictionaries, or tuples. Restored objects can be passed directly to
``get_position()`` and ``add_ohlc()``.
Sets still restore as lists, with their asset values preserved.

Older backups may contain an untagged asset dictionary. Restoration reconstructs
it only when the same variable path already contains an ``Asset`` initialized by
the strategy. Saved contract details remain authoritative; initialization supplies
the expected type, not replacement quantities, strikes, or signals. Scheduled-file
restoration preserves the original bytes in a permission-restricted sibling
``.legacy-<sha256>.bak`` before applying the recovered variables.
Paths without an initialized instrument remain ordinary dictionaries because
the same shape can be strategy metadata. Migrate those explicitly with
``Asset.from_dict(value)`` after verifying the instrument path.
Do not apply this conversion to arbitrary dictionaries or discard other saved
strategy state. Backups written by the updated runtime should be restored by
the updated runtime; older versions do not understand the asset type tag.
For hosted scheduled bots, preserve the original remote state independently
before migration: a local sibling backup is not a durable cloud backup unless
the hosting system explicitly retains it.

FRED Data Errors
--------------------------------------------------------------------------------

Use official FRED series identifiers rather than market symbols. A failed series
can appear under ``errors`` even when other series succeed in the same snapshot.
HTTP and transport errors report a status or exception type without exposing the
API key in the request URL. Correct an invalid series; do not replace working
credentials because one series returns HTTP 400.

Critical Mistakes (Will Break Your Strategy)
--------------------------------------------

Using an Unfinished Crypto Candle as Historical Evidence
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

CCXT candles are timestamped at their opening time. At 09:00, the eventual
close of the 09:00–09:01 candle is not known. Default CCXT backtest history
and last-price queries therefore use completed candles only. This applies to
minute, hour and day bars, including AI research through those methods.

Do not add a negative history ``timeshift`` to make that future closing price
available to your strategy. The backtesting broker uses an execution-only
offset to simulate orders against the current candle; this does not make the
whole candle valid evidence for the preceding decision. When no candle has
closed yet, missing history is expected, not permission to invent a price.

Using datetime.now() Instead of self.get_datetime()
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Backtest results will be completely wrong. Your strategy will think it's the current date instead of the simulated date.

.. code-block:: python

    # WRONG
    current_time = datetime.now()
    target_date = datetime.today() + timedelta(days=30)

    # CORRECT
    current_time = self.get_datetime()
    target_date = self.get_datetime() + timedelta(days=30)

Adding 'from __future__ import annotations'
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Causes immediate crash during backtesting.

.. code-block:: python

    # NEVER DO THIS - WILL CRASH YOUR STRATEGY
    from __future__ import annotations

Simply remove this import. It's not needed and will break everything.

Assigning Attributes Directly on self
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Overrides Lumibot internals, causing crashes or unexpected behavior.

.. code-block:: python

    # WRONG - collides with framework
    def initialize(self):
        self.name = "MyBot"
        self.asset = Asset("SPY")
        self.symbol = "SPY"

    # CORRECT - use self.vars
    def initialize(self):
        self.vars.strategy_label = "MyBot"
        self.vars.target_asset = Asset("SPY", asset_type=Asset.AssetType.STOCK)
        self.vars.target_symbol = "SPY"

Forgetting set_market("24/7") for Crypto
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Crypto bot stops trading at 4pm EST every day.

.. code-block:: python

    # WRONG - crypto bot stops at 4pm
    def initialize(self):
        self.sleeptime = "1M"

    # CORRECT - crypto trades 24/7
    def initialize(self):
        self.set_market("24/7")  # REQUIRED for crypto
        self.sleeptime = "1M"

Data Mistakes
-------------

Not Checking if get_last_price() Returns None
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Crashes with NoneType error when data is unavailable.

.. code-block:: python

    # WRONG - will crash on None
    price = self.get_last_price(asset)
    quantity = self.portfolio_value / price  # Crashes if price is None

    # CORRECT - always check for None
    price = self.get_last_price(asset)
    if price is None:
        self.log_message(f"No price for {asset.symbol}", color="red")
        return
    quantity = self.portfolio_value / price

Using get_historical_prices() for Real-Time Data
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Strategy uses stale data, missing current price movements.

.. code-block:: python

    # WRONG - historical data can be 1 minute delayed
    bars = self.get_historical_prices(asset, 1, "minute")
    current_price = bars.df.iloc[-1]["close"]

    # CORRECT - use get_last_price for real-time
    current_price = self.get_last_price(asset)

    # BEST - use get_quote for bid/ask
    quote = self.get_quote(asset)
    if quote and quote.bid and quote.ask:
        mid_price = quote.mid_price

Returning Early When get_greeks() is None
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Strategy stops running for the entire iteration just because one option has no Greeks.

.. code-block:: python

    # WRONG - blocks entire strategy
    greeks = self.get_greeks(option_asset)
    if greeks is None:
        return  # Strategy stops here!

    # CORRECT - continue with other logic
    greeks = self.get_greeks(option_asset)
    if greeks is not None:
        delta = greeks.get("delta")
        # Use delta here
    else:
        self.log_message("Greeks unavailable, skipping delta check", color="yellow")

    # Strategy continues with other logic...

Options Mistakes
----------------

Manually Selecting Option Expirations
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Selected expiration may not have tradeable data during backtesting.

.. code-block:: python

    # WRONG - expiration may not exist
    target_expiry = self.get_datetime() + timedelta(days=30)
    option = Asset("SPY", asset_type=Asset.AssetType.OPTION,
                   expiration=target_expiry.date(), strike=400, right="call")

    # CORRECT - use OptionsHelper
    chains = self.get_chains(underlying_asset)
    target_expiry = self.get_datetime() + timedelta(days=30)
    valid_expiry = self.options_helper.get_expiration_on_or_after_date(
        target_expiry, chains, "call"
    )

Brute-Forcing Deltas by Scanning Strikes
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Backtests become very slow because calling ``get_greeks()`` repeatedly can trigger many option quote-history downloads.

.. code-block:: python

    # WRONG - brute-force scan (slow)
    strikes = chains.strikes(expiry, "PUT")
    for strike in strikes:
        option = Asset("SPY", asset_type=Asset.AssetType.OPTION, expiration=expiry, strike=strike, right="put")
        greeks = self.get_greeks(option, underlying_price=underlying_price)
        ...

    # CORRECT - use OptionsHelper (bounded probing + caching)
    strike = self.options_helper.find_strike_for_delta(
        underlying_asset=underlying_asset,
        underlying_price=float(underlying_price),
        target_delta=-0.20,
        expiry=expiry,
        right="put",
    )

Forgetting Options are 100x Multiplied
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Position sizing is off by 100x.

.. code-block:: python

    # WRONG - buys 100x too many contracts
    option_price = 1.50  # $1.50 premium
    contracts = 10000 / option_price  # Wrong: 6666 contracts!

    # CORRECT - account for multiplier
    option_price = 1.50
    actual_cost_per_contract = option_price * 100  # $150
    contracts = int(10000 / actual_cost_per_contract)  # Correct: 66 contracts

Using get_last_price() for Options Pricing
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Stale or missing prices for illiquid options.

.. code-block:: python

    # WRONG - last trade can be very stale for options
    price = self.get_last_price(option_asset)

    # CORRECT - use quote for bid/ask
    quote = self.get_quote(option_asset)
    if quote and quote.bid and quote.ask:
        fair_price = quote.mid_price
    else:
        self.log_message("No valid quote for option", color="yellow")

Order Mistakes
--------------

Expecting Immediate Position Updates After submit_order()
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Strategy logic based on outdated position data.

.. code-block:: python

    # WRONG - position hasn't updated yet
    self.submit_order(order)
    position = self.get_position(asset)  # Still shows old data!

    # CORRECT - check on next iteration
    self.submit_order(order)
    # In the NEXT on_trading_iteration():
    position = self.get_position(asset)  # Now updated

Using submit_order() to Close Crypto Futures
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Opens a new position instead of closing existing one.

.. code-block:: python

    # WRONG - opens opposite position instead of closing
    order = self.create_order(futures_asset, quantity, "sell")
    self.submit_order(order)

    # CORRECT - use close_position
    self.close_position(futures_asset)

Using Deprecated take_profit_price/stop_loss_price
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** May cause order errors or unexpected behavior.

.. code-block:: python

    # WRONG - deprecated parameters
    order = self.create_order(asset, 100, "buy",
        take_profit_price=110,
        stop_loss_price=90)

    # CORRECT - use secondary_ parameters
    order = self.create_order(asset, 100, "buy",
        order_class=Order.OrderClass.BRACKET,
        secondary_limit_price=110,      # Take profit
        secondary_stop_price=90)        # Stop loss

Visualization Mistakes
----------------------

Adding Markers Every Iteration
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Chart crashes or becomes unusable due to thousands of markers.

.. code-block:: python

    # WRONG - adds marker every iteration
    def on_trading_iteration(self):
        self.add_marker("Price", price, color="blue")  # Chart explodes!

    # CORRECT - markers only for significant events
    def on_trading_iteration(self):
        if signal_detected:  # Only when something happens
            self.add_marker("Buy Signal", price, color="green", asset=my_asset)

        # Use add_line for continuous data
        self.add_line("SMA_20", sma_value, color="blue", asset=my_asset)

Using 'text' Parameter in add_marker/add_line/add_ohlc
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** TypeError crash - there is no 'text' parameter.

.. code-block:: python

    # WRONG - causes TypeError
    self.add_marker("Signal", price, text="Buy now!")

    # CORRECT - use detail_text for hover text
    self.add_marker("Signal", price, detail_text="Buy signal triggered", asset=my_asset)

Forgetting to Pass asset Parameter
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Indicators appear in separate subplot instead of overlaying price chart.

.. code-block:: python

    # WRONG - indicator in separate subplot
    self.add_line("SMA_20", sma_value, color="blue")

    # CORRECT - overlays on asset's price chart
    self.add_line("SMA_20", sma_value, color="blue", asset=spy_asset)

Code Organization Mistakes
--------------------------

Hardcoding API Keys
~~~~~~~~~~~~~~~~~~~

**Impact:** Security risk and deployment problems.

.. code-block:: python

    # WRONG - hardcoded fallback
    api_key = os.getenv('PERPLEXITY_API_KEY', 'your_api_key_here')

    # CORRECT - None fallback
    api_key = os.getenv('PERPLEXITY_API_KEY')
    if api_key is None:
        self.log_message("PERPLEXITY_API_KEY not set", color="red")

Setting Parameters in Multiple Places
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Confusing, parameters override each other unpredictably.

.. code-block:: python

    # WRONG - parameters set in multiple places
    class MyStrategy(Strategy):
        parameters = {"symbol": "SPY"}

    if __name__ == "__main__":
        strategy = MyStrategy(parameters={"symbol": "AAPL"})  # Which one wins?

    # CORRECT - one place only
    class MyStrategy(Strategy):
        parameters = {
            "symbol": "SPY",
            "period": 20
        }
        # Never override parameters elsewhere

Using try/except to Hide Errors
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Bugs are hidden, making debugging nearly impossible.

.. code-block:: python

    # WRONG - hides real errors
    try:
        price = self.get_last_price(asset)
        quantity = self.portfolio_value / price
    except:
        pass  # What went wrong? No idea!

    # CORRECT - explicit error handling
    price = self.get_last_price(asset)
    if price is None:
        self.log_message(f"No price for {asset.symbol}", color="red")
        return

    quantity = self.portfolio_value / price

Using sleep() in Strategy
~~~~~~~~~~~~~~~~~~~~~~~~~

**Impact:** Blocks the entire bot, preventing important code from running.

.. code-block:: python

    # WRONG - blocks everything
    def on_trading_iteration(self):
        self.submit_order(order)
        time.sleep(5)  # Bot frozen for 5 seconds!

    # CORRECT - check conditions next iteration
    def on_trading_iteration(self):
        if not hasattr(self.vars, "order_time"):
            self.submit_order(order)
            self.vars.order_time = self.get_datetime()
            return

        elapsed = self.get_datetime() - self.vars.order_time
        if elapsed > timedelta(seconds=5):
            # Now do the next step
            pass


IBKR history waits and data-health diagnostics
--------------------------------------------------------------------------------

A provider rate limit is a retryable wait, not evidence that an instrument has
no prices. When the downloader supplies structured rate-limit information,
LumiBot retains the request identity and exposes the provider wait in download
status. Buying more simultaneous quotes does not automatically remove
historical-data pacing restrictions.

Inspect the ``data_health`` field in backtest settings alongside the requested
window and provider error details. An incomplete diagnostic is not an automatic
backtest failure; a legitimate strategy can also produce no trades. Short daily
requests include the full required history and calendar padding. Longer
stock/index windows retain the five-year page cap and backward pagination.