Trade Strategy Perf Evaluation Flashcards
Trade Motivation
- Profit seeking
Active pm seek to capitalize on investment ahead of the market
- rate of alpha decay: detection in alpha once an investment decision has been made
is high when manager needs to trade in shorter time frames
Profit-seeking algorithms use real-time market data to determine which securities to buy and sell
trade agency: refers to how quickly or slowly the order is executed over the trading time horizon
- risk Management / hedge needs
- Portfolio needs to be traded to maintain target risk exposures
- Derivatives may be used to facilitate risk management - Cash flow needs
- trades are caused by investor subscription into, and redemption is out of the fund
- collateral and margin calls could require close-to-immediate liquidation, whereas a fund redemption entails a notice period
- to avoid cash drag, PM may engage in equalization strategy - corp. action, margin calls and index reconstitution
- M&A or spinoff may require portfolio trading or dividend reverting
- margin calls on leveraged positions that suffering losses may require urgent sale of portfolio holdings
- when the bank index is reconstituted, PM needed to executed trades to reflect the changes
trading strategy inputs
Side
if price increase, executing a buy order may take longer than executing a sell order
- herding effect
Size
the total absolute amount of the security being transacted. Large order have increase market impact cost.
Relative size
order size as a % of the security’s average daily volume (ADV)
ST alpha
alpha decay high
high urgency
increase rates of alpha decay require faster trading and associated higher trading costs
high alpha decay, high urgency
higher market impact cost
Greater trade urgency results in lower execution risk because the order is executed over a shorter period of time, which decreases the time the trade is exposed to price volatility and changing market conditions.
price volatility
increase price volatility implies increase execution risk
security liquidity
increase liquidity reduces execution risk and trading costs such as market impact
market crisis
market volatility and liquidity are negatively correlated
Liquidity is a key concern
Stock Index
During the normal market environment, a firm being added or removed from a stock index may directly affect its liquidity
High risk aversion
high urgency to trade
- more concentrated about market risk
- tend to trade with greater trade urgency to avoid the greater market exposure
market impact (size)
the adverse prices impact caused true trading a large volume order at one time.
To minimize info leakage, pM may choose trading in the dark pool or using market and limit orders.
execution risk
trading too slowly
the adverse price impact resulting from a change in the fundamental value of the security
trader’s dilemma:
all equity market impact causes execution risk and vice versa
tradeoff: urgency rise, market impact increase, execution risk decrease
urgency decline, market impact decline, execution risk increase
Reference prices
used to determine expected trading costs which enables managers/ traders to select the optional strategy for a trade
Pretrade benchmarks (before the start of trading)
A pre-trade benchmark is often specified by portfolio managers who are buying or selling securities seeking short-term alpha by buying undervalued or selling overvalued securities in the market.
decision price
previous close
use as proxy for decision price when it’s quantitative manager
opening price
subjective manager
arrival price: time the order is sent to the market for execution
intraday benchmark
based on prices during trading period, used by managers who trade passively over a day, or funds that maybe rebalancing or mining risk
VWAP
low order volume relative to avg. daily volume traded
narrow bid-ask
low urgency to complete
help achieve the objective of using the cash received from sell order to fund buy orders of the rebalancing
-equal weighted average price of all trades executed
Time-series average price
TWAP
average price ignore volume of all trade during trading horizon
best when there are outliers or if there is highly fluctuating volume during the day
appropriate for managers who wish to remove the impact of outliers since they believe they are less able to participate in extreme trades. It’s also appropriate in market environments with highly fluctuating volume throughout the day.
Post trade benchmark
determine after trading has been completed.
close-price used to reduce tracking error
Price target benchmark
base on manager’s view of fair value
Price used to profit-seeking managers aiming to earn short term alpha, related to the manager’s view of the fair value of the security
Trading Strategy Selection
ST alpha
trade ST mispricing in a liquid market
high urgency
price target Benchmark
LT alpha
trade over the LT due to change in fundamental conditions
low urgency
difficult to use
Risk rebalance
rebalance or hedge risk exposure
low urgency
TWAP
client redemption
liquidate the holdings to meet client redemption
EOD trading close price
New mandate
invest new client fund
low urgency
close price
trade implementation choices
high-touch approach
human
require for large trades (block trade), involving greater human engagement
low order volume relative to avg. daily volume
narrow bid-ask
high urgency
dealer market include: principle trade
dealer/ market makers assume all or some of the risk relating to executing the order, priced into spread.
Quarter-driven, OTC or exchange markets are primacy principal trade markets which also includes request-for-quote (RFQ) markets where market makers do not provide continuous quotes, but only does on request.
- broker market
agency trades dealers try to arrange trades by acting as agents (brokers) on behalf of client.
electronic trading
used for more liquid trades/market
involve direct market access (DMA) and/or algorithmic trading.
DMA allow buy side portfolio managers/ trades to access the order book of the exchange directly through a broker’s tech infrastructure.
algorithm trading
profit seeking & trade execution
Profit-seeking algorithms use real-time market data to determine which securities to buy and sell
execution algorithms focus on how to trade manager orders.
scheduled algorithms % of volume (POV)
Vwap
twap
POV (participation algorithm)
send orders according to volume participation schedule
adv:
automatically exploit increased liquidity when available
disadvantage:
they continue to trade at any potentially adverse price, and may not fill the order in a specific time if there’s a lack of trading
trading cost increase
VWAP & TWAP
time-slicing algorithms
VWAP algorithms attempt to watch the vwap price for the period by carving up the trade and steady slice orders based in historical intraday volumes
adv:
ensure that a specific # of shares are executed in a specific time period
disadvantage:
they may force trades in times of low liquidity or trade too little in times of high liquidity
liquidity-seeking algorithms
opportunistic algorithms
if the trade is high ADV of the daily volume
arrival price
price when sent order to market
algorithms seek to trade close to market prices prevailing at the time the order is entered. Trade more aggressively
Dark strategies / liquidity aggregators
execute trades in dark pools, with aggregate algorithms attempting to optimize trading across multiple farm revenues
Dark aggregator algorithms appropriate for liquid or wide bid-ask spread, large order size. Good for avoid information leakage.
Smart order routers (SORs)
that determine the best destination to route an electronic order to get the best result.
Focus on getting the best price for market orders or the highest probability of execution for limited orders.
Clustering
ML technique whereby a computer learns to identify which algorithm is optimal for different types of traders based on the key features of trades.
High frequency makes forecast
attempt to model ST market direction.
LASSO is a ML technique that helps to reduce the # of explanatory variables to manageable # of significant variable
Equity
Stock exchange
Dark pools
most tech advanced
electronically executed
FI
low liquidity
large order size - dealer based quote driven market
High frequency execution for urgent trades, less urgent trade use broker-agent approach
OTC treasury can be traded algorithm
RFP
Exchange trade derivative
most common electronic trading
using low touch approach
buy-side traders generally use DMA for small trade
CTC Derivatives
dealer quote driven market
high touch approaches
Spot fx currency
OTcmarket both electronic trading and high touch broker approaches
Small order use DMA
Explicit
commission and fess
implicit embedded
market impact/ execution risk
Implementation shortfall IS
total cost =
+ (arrival- decision price) * filled size
+ (Pclose - decision) * unfilled size
+ fixed cost
IP
= execution (delay +trading cost)
+ opportunity
+ fixed fees
=paper - actual
= total shares * (close - decision)
-actual filled (close - execution)
-fixed cost
iS bps = IS$/(decision price * total shares)
execution cost
= delay cost + opportunity cost
= (order executed price - decision price) * order size filled
delay cost= (arrival- decision price) * filled size
opportunity cost = = unfilled shares * (closing price - decision price)
due to executing shares at a less favorable price than the decision price
- delay cost due to adverse price movement intake order submission and trade reduced
minimize the efficient trading practices to give trader pre trade and post trade analysis to make swift decision on best trading strategy.
bps execution cost
execution/decision - 1
* filled order / total shares * 10,000
Opportunity cost
= unfilled shares * (closing price - decision price)
cost of unearned return by the unfilled order
fixed fees
any explicit commissions, or fees, incurred in executing the trade
improving evolution performance
- delay cost can be reduced by having a process in place that provides traders with broker per ethics
- opportunity cost can be minimized by investing in the sub-optimal assets with detailed analysis
absolute cost ($)= GEODE trade slippage
side * (execution price - bmk price) * shares executed
trade cost (bps) =
side * (execution price / bmk price -1) * 10k
Market-adjusted cost (bps)=Arrival cost (bps)−β×Index cost (bps)
=side * (execution/arrival price ) * 10k
- beta *
Side×(Index VWAP/Index arrival -1)×10k
Index cost (bps)= Side×(Index VWAP/Index arrival -1)×10k
ensure a trade is not penalized or rewarded for general market movements over the trade horizon by subtracting the index cost adjustments for security’s beta
if >0
market-adjusted cost is thus significantly lower than the total arrival cost. This indicates that most of the expense associated with buying is due to the effect of buying it in a rising market as opposed to the buying pressure induced by the order itself.
index cost (bps)
side * ( index vwap / index arrival price -1)*10k
added value (bps)
= arrival cost (bps) - est. pre-trade cost(bps)
negative cost = benefit
trade governance
trade policy:
- meaning of best execution
- factors determine the optional order execution approach
- handling trading errors (collation of pm, risk manager ,legal/ compliance)
- listing of eligible broker & execution venues
- process to monitor, execution arrangements
factors determine optimal execution approach
- urgency and size of order
- liquidity of security (APV) and the nature of security
- characteristics of available execution venues
- investment strategy objectives (LT vs. ST)
- Reason for the trade
Performance measurement
provides an overall indication of the portfolio’s performance, typically related to a benchmark
performance attribution
held to understand return, both relative and absolute returns
performance appraisal
determine whether the perf has affected primarily by investment decision, by the overall market, or by chance/ luck.
draw conclusion the quality of pm’s investment decisions.
combine output from both performance measurement and performance attribution to refer a professional judgement on the quality of performance.
effective profess of PA:
- a reflection of 100% of the portfolio’s return or risk exposure
- the portfolio manager’s current decision-making process
- the active investment decisions taken by the pm
- a full explanation of the portfolio’s excess return and risk
must account for 100% of the portfolio’s risk and return and adequately represent the pm’s current investment process
return/ risk attribution
return- based
adv:
- easy to implement
- does not require holdings data useful for private fund (HF)
disadvantage:
- least accurate because does not consider holdings
- return data can be manupulated
risk-based
- parallel of return attribution but analyze the impact of pm’s active management decisions on portfolio risk
holdings-based
adv:
- more accurate than return-based because considers underlying holding
disadvantage:
- needs data on fund holdings
- ignores transaction impact
transaction-based
adv:
most accurate method
disadvantage:
- highest data requirement
- highest complexity
- time consuming to implement
- use both the holding and transaction
- most accurate method but most difficult and time consuming to implement
- underlying data must be complete, accurate, and reconciled from period to period
Micro
analyze the portfolio at the portfolio manager’s level and sees to verify that the PM did what they said they would and to understand the drives of the portfolio’s return
Macro
analyze investment decision at the fund sponsor’s level. Common used with institutional investing quantifies fund sponsor’s decisions to deviate from heir strategic asset allocation and the timing when they made those decisions.
Factor-based return attribution
carport model based on
- value growth index RMRF
- market cap SMB
- book value to price HML
- momentum VWL
allow practitioner to remove the effects of known market factors to quantify the excess return.
arithmetic attribution
focus on active return of a portfolio in a single period
alpha = rp-rb
geometric attribution
periodic excess return of the portfolio
(1+rb)(1+ag) =1+rp
alpha g = (1+rp)/(1+rb)-1
Bronson model
BHB Bronson-hood-beebover
allocation: Ai = (wi-W)* Bi
selection = Si = W (ri-bi)
interaction i= (wi-W)*(Ri-Bi)
BH Bronson-fachler method
allocation: Ai = (wi-W)(BI-BT)
selection = Si =W(ri-bi)
INTERACTION = (w-W)*(Ri-bi)
address draws back of BHB’s allocation effect for segment because the sign of the resulting allocation effect does not automatically indicate whether the decision to oer/underweight a particular segment of the portfolio is correct
in BF, allocation effect looks at the active weight in the segment multiplied by the passive benchmark return of the segment relative to the overall benchmark.
FI attribution
3 common method
- exposure decomposition duration based
- yield curve decomposition -duration based
- yield curve decomposition - full repricing
exposure decomposition duration based
allow single presentation of the output of the attribution with relative low data requirements
- duraion
- curve sharpe
sectorselection - bond selection
yield curve decomposition duration based
- require more data
- more operational complexity
- yield or income
- roll
- shift
- shape
- spread
- residual
yield curve decomposition - full repricing
- most comprehensive
break into individual spot rates for CF occurring at different maturities - most precise and can accommodate the broadest range of instruments and yield curve change
-most complex, least likely to be easily understood
bottom up
relative:
security marginal contribution to tracking risk TE
absolute:
security’s marginal contritbution to total risk
top-down risk based method:
relative:
attribute TE to relative allocation and selection
absolute:
factor’s marginal contribution to total risk and specific risk
factor based
relative:
factor’s marginal contribution to TE and active specific risk
absolute:
factor’s marginal contribution to total risk and specific risk
liability-based benchmarks
- likely to be used by fund sponsors and PMs when a firm has a specific liability to pay in the future
- focus in CF necessary to satisfy the liability and frequently limits the investment choices available to PM
- nominal bonds, inflation-adjusted bonds and high quality stocks
plan factors that impact liabiltiy and PM aa to most CFs:
- # years work on average
- # workers retired
- impact of inflation on the liabilities
- correlation between EDIT & PA
- if plan is frozen or has a terminal life
- avg actuarial assumption, including life expectancy and required discount rate over the plan
Asset-based benchmark 7
- absolute return benchmark
- broad market indexes (sp500)
- Style indexes (large cap value/growth)
- factor model based CAPM
- return-based benchmark
- manager universe (peer group)
- Custom security-based benchmark
- absolute return bmk
a return objective that arises to exceed a minimum target return
adv: simple and straightforward
disadvantage: its’ not an investable benchmark
- broad market indexes
adv:
- well recognized, easy to understand by clients, and widely available
- unambiguous, generally investable, measurable and may be specific in advance
- appropriate to use if it reflects the current investment process to the manager
disadvantage:
- manager’s style may deviate from the style reflect in the index
- differing definition of inv. style can provide diff. bm returns
- style indexes
adv:
- widely available, widely understood by clients, widely accepted
- if the index reflects the manager’s style and it’s investable, it’s an appropriate benchmark
disadvantage:
- some style indexes can contain weightings in certain securities and sectors that may be larger than considered prudent.
- Differing definitions of inv. style can produce quote different benchmark returns, making them appropriate benchmark
- factor model based benchmark CAPM
ADV:
- useful in perf evaluation
- provides managers and sponsors with insight into manager’s style by capturing factor exposure that affect an account’s performance
disadvantage:
- focus on factor exposures is not intuitive to all managers or sponsors
- the data & modeling are not always available and maybe expensive to obtain
- it maybe ambiguous because diff. factor models can produce different outputs, leading to misspecification
- return based benchmark
- managed account returns over specific periods and
- corresponding returns on several style indexes for the same periods
adv:
- general easy to use and intuitive
- meets the criteria of a valid benchmark
- useful where the only info available is account returns
disadvantage:
- the style indexes may not reflect what the manager owns or what the manager or client would be willing to own
- enough monthly returns would be needed to establish a statistically reliable pattern of style exposures
- will not work when applied to managers who change style
- manager universe / peer group
the median pm/fund from a broad universe of manager of fund is used as the benchmark.
Median is the fund that falls at the middle when ranked from the highest to lowest by performace
adv: measurable
disadvantage:
- subject to survivor bias as underperforming managers’ performance results are revamped from the universe history
- rely on the university has been accurate complied
- can not be identified in advanced, so it’s not investable, thus not an acceptable benchmark
- custom security based benchmark
designed to reflect the manager’s security allocation and inv. process
adv:
- meet all required benchmark properties and all the benchmark validity
- allow continual monitoring of inv. process
- allow fund sponsors to effectively allocate risk across inv. management teams
disadvantage:
- it can be expensive to construct and maintain
- a lack of transparency by the manager (HF)
can make it impossible to construct such a benchmark
valid benchmark process:
- specified in advance
know to both pm and fund sponsor at the start of an evaluation period - appropriate benchmark is consistent with the manager investment style
- measurable value & return can be determined on a reasonable frequent basis
- unambiguous
clearly defined identifies and weights of securities - reflective of the manger’s current investment opinion
- accountable
- inevitable
possible to replicate the benchmark and forgo active management
benchmark quality evaluation P=
M+S+A
return on market index + excess return to style + active return
s= B-M style index - market index A= p-B= portfolio return - style benchmark return
rp=
r(market)+r(Style)+r(active)
B Benchmark= r(market )+r(style)
E Expected return= r(style) + r(active)
E=p-m=s+a
A good benchmark should reflect
correlation (S,A)=0 style active
correlation (S,E) OR (S,S+A) >0 style expected return
hedge funds benchmark
- broad market index (Not appropriate)
- rf rate
+ spread unsuitable - hedge fund peer universe
unstable due to survivorship bias and backfire bias, smoothing effect decrease STD, increase sharpe ratio and allocation fo hF
real estate
benchmark exists but not all suitable
- use to appraisal data lead to smoothing effect and
- benchmark return are self-reported so subjective and bias maybe presented
- benchmark are derived from a sample of universe not representative of asset class
- perf. of index bears a very high correction to largest investment
- benchmark that are value-based could be biased toward the most expensive properties or geographical areas
understated volatility or risk - lack of comparability with benchmark return give that same benchmark use leverage while others do not
- indexes assume no transaction cost, full transparency, and normal liquidity which is rally not the case; those factors would impact actual real estate returns
Private Equity
valuation varies by PM
IRR is influenced by target CF at the begging of investment
commodity
basis risk
managed derivatives/ futures fund
survivorship bias
distressed securities
iliquidity and lack of marketbility
stable pricing
appropriate choice of benchmark
garbage in garbage out
perf appraisal
sharpe ratio
= ra-rf /std
measure of additional / excess retrun for bearing risk above the rf rate, stated for unit of return volatility
draw back:
STD does not differentiate b/w volatility that is upside vs. downside.
there’s a penalty for all volatility, even if it’s good volatility
Treynor ratio
systematic risk
Ta= (ra-rf)/(beta)
only considers systematic risk rather than total risk
only appropriate for assume efficient market
only useful for well diversified /systematic risk and do not have unsystematic risk
IR
active return /active risk
appraisal ratio AR
=alpha return (factor based regression )
/ std of error
Alpha is excess return, which is calculated as the return earned by the portfolio minus the return suggested by CAPM
= return - capm return
std of error = sqr root of ( std^2- b^2*(bmk std^2))
Sortino ratio
only target semi-std measure of downside risk
SRP = (E(rp)-rT) / std
rt= minimum accepted return
the larger the ratio, the better the risk adjusted performance
- more appropriate for non-normal (nonsytematical skewed) return distribution
- position & negative skewed investment strategy
result in lower sharpe ratio
capture ratio CR
determine the manager’s relative performance when market are up or down
UC/DC = rp/rb
uc > 1 outperform the market
dc <1 outperform the market
CR= UC/DC >1 postive asymmetry convex
<1 negative asymmetry concave
Maximum drawdown
is the cumulative peak-to-trough loss during a continuous period.
drawdown duration = total time from the start of the drawdown until the cumulative drawdown returns to zero
Drawdown duration is the total time from the start of the drawdown until the cumulative drawdown RECOVERS to zero, which can be segmented into the drawdown phase (start to trough) and the recovery phase (trough to zero cumulative return).
Scheduled algorithms
- do not have expectations for adverse price movement during the trade horizon.
- greater risk tolerance for longer execution time periods and
- more concerned with minimizing market impact.
- often appropriate when the order size is relatively small (e.g., no more than 5%–10% of expected volume)
- the security is relatively liquid, or the orders are part of a risk-balanced basket and trading all orders at a similar pace will maintain the risk balance.
arrival cost(Bps)
= (avg. EXECUTION price / arrival price order sent for execution) -1 *10K
turnover
lower of purchases and sales divided by average monthly net asset.
