Introduction to Stochastic Calculus & Application in Finance

674 回覆
359 Like 8 Dislike
2018-11-05 03:55:35
小弟不才真係唔熟time series 應該答你唔到了
我相信在場大把有料過我嘅巴打都可以幫手答

至於model兒唔兒戲其實就真係兩睇
我地build一個model出黎其實都係想描述一樣係現實世界發生緊嘅嘢
一個model冇話絕對啱定錯 就算係一個非常簡單嘅model都可能好有用
所以到最尾其實都係要睇你信唔信個model
你信即係代表你認為呢個model已經盡可能capture曬個樣事物係現實世界嘅性質
好似你用ARIMA model咁 如果你相信現實世界嘅stock price係有Auto regressive, Moving average同integrated呢啲性質
而你又認為呢個model已經盡可能capture曬stock price嘅性質 (至少係佢最核心嘅性質)
咁ARIMA model對你黎講已經係一個足夠好嘅estimate
正如以前嘅人覺得用newton mechanics去描述地球上嘅物理已經好足夠
但係後來先發現如果要in general描述唔同情況下嘅物理 咁general relativity會係一個比較好嘅model
(如果你問我 我就寧願信time series analysis
都唔會信跌穿唔知邊條線就升 有啲唔知咩pattern就跌 之類嘅嘢 )

同埋有樣嘢我都想搭單問下
乜GARCH model唔係通常用黎model volatility嘅咩
定其實可以直接拎黎model stock price
2018-11-05 09:09:31
巴打見解好有用
通常我地要做到金融既forecasting都唔會淨係用一個model
GARCH會係其中一個 有機會會用到EWMA ARIMA呢類
詳細我都唔清楚 我都只係學緊
2018-11-05 17:59:54
好似有paper 指出Garch 係forecasting volatility and return on volatility. 用落pricing 好似冇印象見過
利申:只係印象中有呢句出現,小弟比起呢到既ching 不才純粹分享
2018-11-05 21:33:36
when pde is available, it is always using pde.
mc is too slow for single dimension case
2018-11-05 23:32:59
玩埋math4210啊嘛
2018-11-07 00:51:03
3.) Black-Scholes-Merton Model
(d) Application of risk-neutral pricing formula

同之前講嘅一樣啦 喺再落下一隻derivative --- bond option之前 首先我地要知道點搵bond price先

希望大家本身都對bond有最基本嘅認知 如果想好快咁重溫一下就睇下圖啦

簡單咁講 每一張bond都會講定一個face value同埋coupon rate
(For simplicity呢個example我假設coupon rate係fixed嘅 當然現實就唔一定係fixed)
然後只要你比錢買(依家假設你係t=0買入) 呢張bond嘅issuer就會每一期都比返coupon你
而coupon最簡單嘅計法就係= Face value * coupon rate (當然可能仲有其他更複雜嘅計法)
當呢張bond到期嘅時候 你就會拎埋最後一期嘅coupon同埋face value
所以換句話說其實你買bond就等於借錢比個issuer

良心建議: 希望冇fina底嘅各位 起碼花少少時間搞明乜嘢係bond先 之前講嘅其他嘢唔明都唔緊要
因為除咗股票 一個正常人最有機會買到嘅instrument就係bond
就算你只係擺錢入銀行做定期其實都已經相當於買咗一張bond

而bond亦都算係比較安全嘅instrument (除非你係專揀啲high yield bond黎買嘅癲佬咁我冇嘢講 )
因為bond issuer原則上係一定要比coupon/face value你 如果佢比唔到就係違約
所以如果你揀啲credit rating好嘅bond issuer去買 基本上係非常非常安全嘅
亦都因為咁所以每個人嘅portfolio都最好有一定portion嘅bond渣係手 就算經濟轉差都唔至於一鋪清袋


咁review完就正式開始
首先我地跟住落黎只會處理zero-coupon bond 即係完全冇coupon嘅bond
而for simplicity我會assume埋face value = 1
佢嘅strucutre請睇下圖

大家都見到真係非常簡單
但係我地只要處理咗最簡單嘅case 咁之後有coupon同face value>1嘅情況其實都好易可以generalize上去
所以我地集中火力處理咗zero-coupon bond先

首先我地諗下year 1嘅finance course係點教我地price呢隻zero-coupon bond先
假設我地依家係continuous compounding啦 咁個result其實就好似下圖咁簡單

大家可以見到zero-coupon bond嘅價錢其實就係個discount factor (depends on interest rate r)
所以當我地問「點樣price一隻zero-coupon bond?」嘅時候
其實變相即係問緊「應該點樣model interest rate?」
而上圖呢個pricing method嘅大前提係interest rate係T-t呢段時間係constant or deterministic
我之前都提過呢個assumption係唔realistic嘅 因為interest rate本質係random
We can do better than this
正如我地之前用GBM去model stock price
跟住我都會介紹一個stochastic嘅interest rate model 然後用佢黎model interest rate

而呢一個model嘅名就係Vasicek model
(CIR model我應該會係section 4尾聲再提 所以請大家等陣啦)
呢個model嘅SDE其實就好似下圖咁

( 請注意 呢條SDE已經係risk-neutral world Q入面嘅dynamics )
我地可以見到呢個model嘅drift term同GBM都幾唔同 不過都唔算係複雜
首先乜嘢係b? 大家可以當b就係interest rate嘅long term mean
而a就係interest rate返翻去呢個mean嘅rate
最尾volatility σ我地for simplicity暫時都係當佢constant先
所以單靠呢條SDE我地其實已經implicit咁假設咗interest rate嘅movement係一個mean reverting process
何謂mean reverting? 就係當你比interest rate亂咁郁一段好長嘅時間 最尾佢一定會返翻去自己嘅long term mean

大家可能會問點解我地要整條唔同嘅SDE出黎玩自己 然後又imply interest rate係mean reverting
我地梗係有原因先咁做 因為有好多empirical study都指出interest rate的確係會revert to mean 所以我地先會用一個咁嘅model去嘗試描述有呢個性質嘅interest rate
(其實呢種process真正嘅名係叫做Ornstein–Uhlenbeck process 有興趣就自己wiki )
咁當然我又唔係話Vasicek model已經係最準確嘅model (大家見到volatility又係constant就知準極有限 )
不過係我地呢個程度 Vasicek model係比較簡單同容易去理解
兼且佢已經足夠幫我地去解釋分析好多嘢 所以大家暫時同住佢玩遊戲先

依家有咗個model就可以返翻去原先嘅問題
假設interest rate follows Vasicek model嘅dynamics
咁一隻zero-coupon bond at time t嘅fair price應該係乜?
我想大家再認真思考下我上面講過嘅呢條式

上面我提過呢隻zero-coupon bond最後其實都係depends on r (當然仲會depends on t)
換個角度諗 其實我地大可以當bond嘅underlying就係interest rate 正如equity call嘅underlying係stock咁
所以我地亦可以大約將bond睇成係一種interest rate derivative
心水清嘅朋友睇我左講右講無論點都要講到"derivative"呢隻字出黎都應該估到我用意如何
我地一路學咁耐derivative pricing 最核心嘅concept係乜?
冇錯就係risk-neutral pricing (如果你唔知我講緊乜就代表你真係要追post )
如果我地將bond當成一隻underlying係interest rate嘅derivative咁睇 咁成個問題就變到非常簡單

我地首先define清楚啲notation先
我會用P(t,T)去表示一隻係time T 到期嘅zero-coupon bond, at time t嘅fair price
咁好自然P(T,T) = 1 因為我已經assume咗face value = 1 所以隻bond到期只會拎返$1
根據risk-neutral pricing嘅神髓 我地會知道P(t,T) = the fair price of zero-coupon bond at time t其實就等於下圖咁樣


大家又再諗下 我地之前assume stock price follows GBM嘅時候
我地係assume曬r同σ都係constant
所以exp[-r(T-t)] 係一個(up to information t) constant 我地其實係因為咁所以先可以將呢個discount rate寫係expectation出面
咁依家interest rate已經係stochastic我地就冇得偷雞再咁做
不過其實呢舊discount rate嘅form唔難搵到出黎亦都唔難去理解
但係個問題係我又寫到爆字數
下個cm再戰 (預告:一個好耐冇出現嘅F字頭朋友會再次登場 )
2018-11-07 01:59:00
少少comments?
1. 圖片公式中的discount rate 應該係discount factor
2. 上面highlight左 原則上 & 違約 等字眼, 但你個analysis係assume default-free bond
3. 其實可以用standard notation exp(-int_t^T r_s ds)
4. 通常對Vasicek的最大的批評係 it allows for negative interest rate
2018-11-07 02:08:24
1. 呢個真係我打錯 呢幾日精神都好差請見諒

2. 我話原則上唔違約嗰段係泛指緊現實世界嘅bond
當然你講得冇錯我之後analyze嘅bond都係default-free 如果唔係包埋credit risk就太複雜
可能我表達得唔係咁好

3. 其實我就係想特登避開呢個notion 等到下個cm先講
因為普通人其實係唔會知discount factor可以咁寫 我想慢慢由constant interest rate嗰個寫法一步一步黎
等讀者可以思考下點解去到stochastic/deterministic嘅時候要用integration

4.經過呢十年呢個批評已經自動invalid啦
2018-11-11 00:20:12
LM
2018-11-12 17:45:31
Updates?
2018-11-12 17:48:38
小弟應該要11月尾先可以再update
依家整緊一個好重workload嘅project
應該暫時都冇時間出文
2018-11-12 17:57:47
No problems. We can just chitchat. What project is it?
2018-11-12 18:00:22
Pricing exotic derivatives using SV+Jump model
2018-11-12 18:00:55
Heston Model?
2018-11-12 18:03:11
Yes we are using heston for the SV part
2018-11-12 18:04:18
I had a paper on Heston Model
2018-11-12 18:10:35
Would u mind sharing the paper with us?
2018-11-12 18:14:12
It’s not published yet and I’m a co-author so I don’t know if I can just share the manuscript. But I can tell you as much as I can.

It’s on the path simulation of the Heston model. Is your project about path simulation for exotic derivatives?
2018-11-12 18:44:24
The paper is about efficient simulation of the Heston model. Simulating the paths of Heston model is essentially simulating the volatility process. Our paper proposed a path discretisation scheme which is a random walk with a changing walk size.
2018-11-12 19:45:06
份notes好似咁4001咁嘅
2018-11-12 20:00:17
We’ve decided to use QE discretization to simulate Vol and Stock price since we all know what it is trying to do
Could you plz explain more abt the discretization scheme in your paper?
2018-11-12 20:01:19
係似 不過4007頭兩堂就教曬4001所有嘢
2018-11-12 20:26:35
未讀到就已經畢業
2018-11-12 22:42:04
QE by Andersen? It’s an efficient method. Our method is actually similar, but faster. Basically we approximate the transition density of the volatility process by a Bernoulli trial.

QE is just an approximation of the noncentral chi square distribution. Its idea is very simple.

Did your lecturer teach QE in the lecture? Or you looked up the paper yourself?
吹水台自選台熱 門最 新手機台時事台政事台World體育台娛樂台動漫台Apps台遊戲台影視台講故台健康台感情台家庭台潮流台美容台上班台財經台房屋台飲食台旅遊台學術台校園台汽車台音樂台創意台硬件台電器台攝影台玩具台寵物台軟件台活動台電訊台直播台站務台黑 洞