威斯康星州拉克罗斯:Quant Strategy for Trading: 面经之八:某人的多次面试总结
来源:百度文库 编辑:中财网 时间:2024/04/28 12:47:22
面经之八:某人的多次面试总结
这段时间光顾着选offer兼灌水了,经常得益于大家的意见,我也写篇面经回馈一下吧。不光讨论面试题,也谈谈经验。有点乱,大家将就着看吧。题库在末尾,前面大半是
我的口水。。。
* DISCLAIMER
以下均个人意见,欢迎指正。请大家千万不要给我站内发信。有什么事情发个帖子让大
家讨论,比私下交流更好。我每次发个有一丁点内容的帖子,就会收到几十封大同小异
的“请教”/“求助”/“问候”贴,实在消化不了啊。其实我是个菜鸟,很菜很菜的那
种。
* 个人背景
本人的背景是MFE,非PhD. 个人感觉这个仍是就业市场上最flexible的配置。从获得面
试的容易程度来说,一页纸(正面)骗了无数面试,什么方向都有。Sales & Trading
Analyst/Associate, Sell-side quant, buy-side strategiest,trading firm FE等
等。得益于学校的营销,什么地方的也都有,伦敦,百慕大,多伦多,香港,纽约,得
克萨斯,加州等等。当然成不成另当别论,面试的机会是不缺的。学校career service
很重要。
* 面试的质量vs数量
简单来说当然是越多越好。本人是在职找工作,时间太稀缺。没时间系统准备面试,刚
开始甚至连简历都没什么时间投。隔三差五的请假面试,外地的,包括方向不是很喜欢
的。很容易把驴丢了,马也没找到。后来就改变策略,全凭个人喜好精心挑选职位,据
掉一切如鸡肋公司,及鸡肋职位,把时间用到你愿意接受的职位上。如果是全职学生,
那么就需要更好的定位,毕竟没有饭碗的压力是很大的,这个时候可以多多面试,以赛
代练。这些好像都是废话。不过这个我觉得对我个人来说是拿到offer的关键,时间有
限,好钢用到刀刃上。
* 背景硬伤:once IT, always IT?
这个绝对是我听过的最无聊的狗屎言论。如果你在找Front Office工作,你只要明白一
点,那就是这类职位本身是非常困难的,如果你失败了,基本上是因为你自己的不足,
跟任何形式的背景歧视无关。举例来说,在我短短的职业生涯里,见过的programmer转
trader(flow or prop)的,甚至后来做掌门的,绝对多于quant转trader的。我认为
原因只有一个,那么就是programmer比quant多,人口基数大,无他。早期入行做Quant
的,之前也都是程序员。Full-time的同学里找工作也是五花八门,之前大多都是各种
类型的programmer.(again,程序员人群太壮大了)
* 猎头陷阱
接着上面讲。如果你之前是IT背景,想转方向花钱弄了个MFE什么的,然后发现自己毕
业后重投IT了,除了实力问题,基本上是脑子糊涂被猎头忽悠了。愚以为猎头对有经验
的candidate的帮助是巨大的。比如从挑选职位,初期申请等工作上,自己可以节约非
常多的时间。Quant的就业市场里,PhD也可以视为这一类。如果你原来是写code,或者
是种大麻的,跟猎头打交道的后果就是你接着写code或种大麻。当然Quant也是写code
的,我们要认清楚这一点。
大家在这点上可以参考Wilmott论坛御用皮条客Dominic Conner的言论。我认为此君虽
然狗屎NC言论诸多,但是在猎头陷阱这个问题上一语中的。猎头最喜欢拿Quant
Developer之类的头衔来忽悠人。检验所有Quant职位的试金石是:reporting line,
bonus pool. 现在连什么credit/market risk IT的HR都会拿这个骗人了,所以一定要
小心。碰到这种电话,问完以上两个问题,如果不对马上闪开。如果去面试了,发现只
问C++,最后拿到offer了,这个也坚决不能要。
不要想着先进门然后内部转职位。这个太难了。你干的太好了,老板绝对不放你。干得
太差了,review不好也没戏。
* C++/编程能力
这个绝对重要。综合以前帖子里谈到的,buy side分工不是很细,什么都要会一点,要
有能够独挡一面的编程能力。就算是bank里的junior quant,编程能力是过关的法宝。
没有这个,根本就过不了前几关。现在这个job market,基本上就是先用几轮c++/编程
面试/brainbench刷掉一半人,以前很多人背几句C++,不会编程也拿到offer了,现在没
有这样的好事了。我印象是前几年大多流行算法题,C++ code test.现在我居然还经常
碰到语言细节茴香豆题。
* MFE选校和找工作
上面谈到的猎头背景和的IT背景硬伤的问题关系巨大.如果不是PhD,怎么找工作?自己
投+学校career service。为了避免这个帖子被口水淹没,我只说一句:远离烂校,远
离没有就业支持的牛校烂MFE项目,非PhD远离猎头。
我的感觉是很多公司已经开始消减开支,我经常直接收到公司内部HR电话,都希望直接
绕过第三方。建议随便投几个公司网站上的职位,这样简历就入数据库了,以后就经常
会有HR直接联系你面试各种职位。如果不满意,好好跟人说没兴趣,让他们接着找。
***********************************************************
***********************************************************
* 面试题库
我一般潜水直接跳过任何面试题的讨论,但是我还是把我遇到过,不是很烂大街,还记
得住的题目,概念列一列。题目以下粗略分类:
Math, Prob/Statistics, Algorithms, Products, S&T/MBA类话题
简单谈谈面试技巧
- Think out loud.这是唯一在面试中取得partial credit的办法。这也是挫人拿offer
的不二法门。
- 知之为知之,不知为不知。但是你可以说,这个东西我不知道,但是,try me, I
might figure out.
- 主动要求提示。这个跟第一点是一个意思,告诉面试管你在哪里stuck了,要个提示
接着做下去。
- 察言观色,不要得罪人。
最后一点要简单举例说说。有的PhD面试官比较subjective,比较容易对你的观点,方法
表示异议。先听听他们的观点,满足一下他们的表现欲,适当附和。答对几道题不是很
重要,最重要的是让对方相信他能够和你一起工作。大部分的MBA/S&T面试官喜欢跑火
车,这点更加重要。我曾经与一个香港大行的eq deriv desk的小美trader讨论三鹿事
件。我系统地列出论据,最后拍拍脑袋总结说三鹿要完蛋了。他不同意,说三鹿的基本
面非常好。凭着对祖国的了解,我忍不住反驳说三鹿的品牌已经完全被摧毁了。我挂掉
了面试,但是几个月后三鹿真的就倒闭了,it当然也不会打电话来请我接着面试。好汉
不吃眼前亏。
*****
Math:
*****
- Ito's lemma. State and derive.
- Brownian Bridge. Eg E[B(0.5)|B(1)]
- Girsanov Thm / Change of measure.
- First passage time of Brownian motion.
- Martingale/stopping time. (Dr. Zhou's book)
- BS derivation. BS PDE as a Heat equation. Homogeneity of S, K. Calculation
of greeks. (google: Efficient computation of option price sensitivities
using homogeneity and other tricks)
- Given f(t,x) satisfies heat equation, given plot of f(t_0, x) vs x,
roughly plot the trajectile of f(t_1, x)
- Monte Carlo: justify monte carlo simulation. (law of large numbers and
central limit theorem.)
- Monte Carlo: integrate f(x) over [a,b] numerically given a random number
generator.
- Monte Carlo: derive the standard error of different variance reduction
techniques(antithetic variables, control variables, etc).
- Solving simple ODE's.
- Linear Algebra: determinant, eigen values/vectors, QR decomposition etc
and which matrices are diagonizable.
- Minimize f(x) = |x-a| + |x-b| + |x-c| + |x-d|
- Solve the GBM/Vacicek/CIR SDE.
- Linear/non-linear curve fitting.
- Numerical analysis: condition of a problem, ill posed problems, error
analysis. know how to formulate problems issues to avoid numerical issues.
eg normal equations in linear regression, beta = (X' * X)^(-1)*X'*y. where
could numerical issues arise etc.
- Optimization - KKT conditions, lagrange multipliers. convexity of
optimization problems.
****************
Prob/Statistics:
****************
- lots of silly Bayes/Total Probability problems. using Bayes twice and
conditioning are usual tricks.
- X1, X2,..., XN iid. prove correlation of X_1^2/sum(X_i^2) and X_2^2/sum(X_
i^2) is negative. things of this sort.
- ant sitting on one corner of a cube and can walk on edges of the cube. it
takes one unit of time to go from one corner to a neighboring corner. it may
go in any direction randomly each time. what is the expected time that the
ant reaches the opposite corner?
- variants of the gambler's ruin problems.
- derive linear regression beta.
- linear regression. Y ~ X. what happens when there are noise in variable X?
- best estimator for variance given absolute values of each sample point.
- cook up dependent random varibles with zero correlation.
- back testing on wrong data. strategy build for max(X_i) input but want to
backtest sum(X_i) data.
- co-integration strategies and tests for co-int.
- CAPM, why estimate we estimate beta with returns, not stock prices?
- Sharpe ratio - various questions. risk/reward. understand what Sharpe is.
Eg. allocate capital to 2 PM's. do you allocate all capital to the PM with
higher Sharpe?
- consistent & unbiased estimates. Hypothesis testing. Error or type I & II.
**********
Algorithms
**********
- sorting and searching.
- Hashing.
- implement a set. complexity of insertion/membership test. (hash/tree)
- data structures in general. lists/maps/vectors. insertion,removal,
iteration, complexity etc.
- N points (x_i, y_i) in a plane(3D space). find circle(sphere) containing
most points. complexity?
- O(N) algorithm for calculation maximum drawdown given daily pnl in an
array.
- byte shifting, multiply by 7. (Dr. Zhou)
- 2 glass balls. building with N floors. glass balls will break if we drop
if from some floor F and above. find F. optimal strategy? complexity?
*********
Products:
*********
- forward vs futures. derive risk neutral pricing formulas.
- Binomial tree - moment matching. (refer to Hull)
- Price a european call by BS and BT. How do you make sure BS and BT prices
are always consistent given BT of N steps?
- Put-call parity: derive arbitrage bounds, maximum call/put price given
infinite volatility etc.
- Delta Hedging: assume volatility follows its own process sigma_t. bought
option at BS price. what are the hedging gain/loss? derive an expression
using black scholes argument.
- Find/derive expression relating spot delta and strike delta of BS call. (
dS * dC/dS = dK * dC/dK)
- compute integral over {0,T} for sigma_t. (fair variance) assuming dS_t = r
* S_t * dt + sigma_t * S * dW_t
- Derive value of european call with non-constant interest rate. (use
forward measure)
- Derive value of log contract. payoff: log(S_T)
- Contract pays $1 when stock price hits $100. price by replication.
- Himalaya option. 3 stocks, pay best performing stock returns at year 1,2,3
. How to price? Correlation exposure?
- Best of two option. pays max(0, (S1_T - S2_T) - K ). Derive price. long/
short correlation?
- given volatility term structure, how to calculate vega given a black blox
premium function f(vol).What can go wrong. (calendar arbitrage etc)
- Give example of non-gaussian mean reverting process for assets/rates. Key
components? (mean reverting term, choice of vol/diffusion process) derive
option/bond prices.
- sticky strike & sticky delta. What is delta under each assumption? Vega?
How to do PnL (Taylor) projection for delta and Vega.
- implied volatiliy surfaces & local vol.
- signs of greeks for call options. greeks for exotics like Asian options,
Variance swaps.
- Digital options. derive black scholes price. price without assuming BS
framework.
- American Options pricing. Monte Carlo? Finite Difference?
- Asian options: pricing with numerical methods and analytical
approximations.
- Swap curve. convert fix rate mortgate to floating rate mortgage etc.
- bootstrapping yield curves.
*******
S&T/MBA
*******
- discounted cash flows. (duh...)
- NPV vs IRR. why is NPV a superior metric.
- ito's lemma. (the typical MBA only knows the formulation in Hull. know
your audience. don't argue.)
- stocks to long/short. pitch.
- different arbitrage strategies. convexity argument.
- be able to hedge every single instrument you talk about. (replication,
dynamic hedging) Posted by kewei tang at 23:48
2 comments:
恭喜,谢谢分享了。我觉得现在经济开始复苏,求职市场也应该会好起来的。我是海脚论坛的一个版主,能否把这篇东西分享过去呢?
http://haijiaobbs.com/forumdisplay.php?fid=6谢谢分享,
我现在想问一个问题,
我拿到了UIUC MFE的AD, 然而面对以后的就业,面对这个新开两年项目的不明朗。有很多的疑惑。能不能帮我解答一下呢?
我的联系邮箱waynezkl@gmail.com
DSP Implementation of Predictive Control Strategy for Power Factor Correction (PFC)应该怎么翻译?
寻找原文书籍:Performance Measurement and Control Systems for Implementing Strategy :Text and Cases
寻找原文书:Performance Measurement and Control Systems for Implementing Strategy :Text and Cases
FOR EVER H.O.T
Market-nicher strategy
什么是communication strategy?
clicks-and-mortar strategy
management strategy 什么意思
mary quant jelly babies是什么意思?
DON,T FORGET ME FOR EVER
DON,T FORGET ME FOR EVER
Haven’t seen you for ages
Don't for get me 是什么意思
英文mary quant jelly babies是什么意思
GRE Quant 和 GRE Analyt 是什么意思?
求助 Don't cry for me ,Argentina 钢琴版
What two things can't you have for breakfast?
What two things can't you have for breakfast?
don't cry for me angentina 歌词?
l for in love lcan t sduty 是什么意思啊
“...you don't know understand .....waitting for you...."
Won't you come in for a moment?
Jim can't wait for the presents,是什么意思?
That child can't keep still for a moment.